Control and Optimization of Fluidized Catalytic Cracking Process

CONTROL AND OPTIMIZATION OF FLUIDIZED CATALYTIC
CRACKING PROCESS
by
PRASAD D. KHANDALEKAR, B.Ch.Eng.
A THESIS
IN
CHEMICAL ENGINEERING
Submitted to the Graduate Faculty

of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
MASTER OF SCIENCE
IN
CHEMICAL ENGINEERING
August, 1993
ACKNOWLEDGEMENTS
I wish to express my sincere appreciation to Prof. James

B. Riggs, chairman of my committee, for his guidance,
encouragement and financial support throughout this study. I
extend my sincere thanks to my other committee members, Dr.

R. R. Rhinehart and Dr. R. w. Tock for their invaluable
suggestions and assistance throughout this work. I also wish
to express my gratitude to Dr. Chang-Bock Chung for his
constant encouragement and valuable advice.
Appreciation is extended to Tammy Kent, Dawn Eastman and
Marybeth Abernathy for their clerical assistance during the
course of this work.
I wish to take this opportunity to thank all my family
members for their love and moral support throughout my career.
I also thank my friends for their support and encouragement.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ii
LIST OF TABLES viii
LIST OF FIGURES ix
CHAPTER
1. INTRODUCTION 1
1.1. Nature of the FCCU Process 1
1.2. Objective of an Ideal FCCU Control System 3
1.3. Scope of Work 4
2. LITERATURE SURVEY 6
2.1. Fluidized Catalytic Cracking Process

and its Control 6
2.1.1. Catalytic Cracking and Coke Formation 6
2.1.2. Regeneration of Catalyst 9

2.1.3. Dynamic Simulation and Control
of FCC Unit 11
2.1.3.1. Dynamic Simulation of FCC Unit 11

2.1.3.2. Control of FCC Unit 13
2.2. Multivariable Process Control 17
2.2.1. Linear Model Based Control 18

2.2.2. Nonlinear Model Based Control 19
3. PROCESS DESCRIPTION AND DYNAMIC SIMULATION 24

3.1. Process Description 24
3.2. Feed Preheat System 25
3.3. Reactor-Riser Model 26
3.3.1. Amoco Model 27
iii
3.3.1.1. Coke and Wet Gas Yield Models 27
3.3.1.2. Reactor Mass Balance 28
3.3.1.3. Reactor Riser Energy Balance 29
3.3.1.4. Reactor Riser Pressure Balance 30
3.3.1.5. Reactor and Main Fractionator
Pressure Balances 30
3.3.2. Lee Model 31
3.4. Wet Gas Compressor 35
3.5. Regenerator 37
3.5.1. Energy Balance 39
3.5.2. Carbon Balance 40
3.5.3. Standpipe Inventory Balance 41
3.5.4. Mass Balance 41
3.5.5. Volume Fraction of Catalyst 43
3.5.6. Catalyst Entrainment 44
3.5.7. Pressure Balance 44
3.5.8. Bed Height 45
3.5.9. Air Lift Calculations 46
3.6. Air Blowers 47
3.6.1. Combustion Air Blower 47
3.6.2. Lift Air Blower 49
3.7. Catalyst Circulation 51
3.8. Numerical Procedure and Validation of Simulator 52
3.9. Analysis of Open-Loop Responses 54
3.9.1. 2% Positive Step Change in Coke
Formation Constant 54
3.9.2. 2% Negative Step Change in Coke
iv
I
3.9.3. 20°F Positive Step Change 10
Fresh Feed Temperature 55
3.9.4. 20°F Negative Step Change I
10
3.9.5. 2% Positive Step Change in Feed Rate 57
3.9.6. 2% Negative Step Change in Feed Rate 57
3.9.7. 30°F Positive Ramp Change in
Ambient Air Temperature 58
4. NONLINEAR PROCESS MODEL BASED CONTROL 98
4.1. Approximate Controller Models 99
4.1.1. Approximate Model for Reactor 100
4.1.1.1. Calculation of Gas
Oil Conversion 101
4.1.2. Approximate Models for the Regenerator 102
4.1.2.1. Approximate Model for Flue Gas
Oxygen Concentration 103
4.1.2.2. Approximate Model for
Regenerator Temperature 109
4.2. Parameterization of an Approximate Model 111
4.2.1. Parameterization of Reactor Model 112
4.2.2. Parameterization of Regenerator Models 113
4.3. Nonlinear Process Model Based Controller 114
4.3.1. Generic Model Control (GMC) Law 114
4.3.1.1. Steady-state Version of GMC
Control Law 116
4.3.2. Flue Gas Oxygen Concentration Control 116
4.3.3. Regenerator Temperature Control 117
4.3.4. Reactor Temperature Control 118

4.4. Basic Regulatory Controllers 118
4.4.1. Riser Feed Temperature Controller 119
4.4.2. Reactor Pressure Controller 119
v
4.4.3. Catalyst Circulation Rate Controller 119
4.4.4. Regenerator Air Flow Rate Controller 120
5. CONSTRAINT CONTROL 121

5.1. FCCU Constraints 121
5.1.1. Bard Equipment Constraints 121
5.1.2. Operating Constraints 122
5.2. Constraint Control 124
5.2.1. Firebox eonstraints 125
5.2.1.1. Approximate Model for Firebox 125
5.2.1.2. Approximate Model for Reactor
Riser 127
5.2.1.3. Calculation of Feed Rate for
Firebox Constraints 127
5.3. Air Flow Rate and Catalyst Circulation
Rate Constraints 128
5.3.1. Approximate Models 128
5.3.1.1. Estimation of Carbon
Concentration on Regenerated
Catalyst 128
Concentration on Spent
Catalyst 129
5.3.1.3. Regenerator Steady-State
Energy Balance 130
5.3.1.4. Reactor Steady-State
Energy Balance 130
5.3.1.5. Overall Steady-State Energy
Balance 131
5.3.1.6. Regenerator Steady-State Oxygen
Balance 132
5.3.2. Parameterization of Approximate Models 132
5.3.3. Calculation of Feed Rate 134
vi
6. FCCU OPTIMIZATION 136
6.1. Development of Objective function 136
6.2. Optimization Models and Parameterization 139
6.2.1. Gasoline Yield Model 139

6.2.2. Carbon Concentration on Catalyst 143
6.2.3. Regenerator Oxygen Balance 143

6.2.4. Steady-State Energy Balances 143
6.2.5. Calculation of Manipulated Variables 143
6•. 3. Optimization Procedure 144
7. RESULTS AND DISCUSSION 146

7.1 Proportional-Integral Controller 146
7.2 Tuning of Control Loops 147
7.3 Unconstrained Control Results 150
7.4 Constraint Control Results 154
7.5 Optimization Studies 156
a. CONCLUSIONS AND RECOMMENDATIONS 190

8.1. Recommendations 191
BIBLIOGRAPHY 194
APPENDICES
A. NOMENCLATURE 197
B. INITIAL CONDITIONS FOR DYNAMIC SIMULATION 209
c. STEADY-STATE DATA FOR PARAMETERIZATION 212
D. DYNAMIC RESPONSES FROM ORIGINAL SIMULATOR 214
vii
LIST OF TABLES
7.1 Tuning Constants for Basic Regulatory

PI Controllers 158
7.2 Tuning Constants for Upper Level

PI Controllers 159
7.3 Tuning Constants for Nonlinear Process
Model Based Controller 160
7.4 Tuning Constants for Constraint Controller 161
7.5 Optimization Results for P • 0.45 $/gal,
P91 = o.52 $/gal, P9 • = o.4£rs/gal 162
7.6 Optimization Results for P 0 = 0.45 $/gal,
P91 = 0.55 $/gal, P9 • = 0.4Ef $/gal 163
7.7 Optimization Results for P~0 = 0.45 $/gal,
P91 = 0.52 $/gal, P9 • = 0.6u $/gal 164
B.1 Initial Conditions for State variables 209
B.2 Initial Setpoints for regulatory Controllers 210
B.3 Initial Valve Positions 211
C.1 Steady-State Feed Pre-Heater data 212
C.2 Reactor-Regenerator Steady-State Data 213
viii
LIST OF FIGURES
2.1 Schematic of Conventional Control Scheme 22

2.2 Schematic of Kurihara Control Scheme 23
3.1 Schematic of FCC Unit 59
3.2 Feed Pre-Beat System 60
3.3 Reactor 61
3.4 Comparison of Reactor Riser Models (Effect of
Catalyst to Oil Ratio on Production of Coke) 62
Feed Temperature on Riser Temperature) 63
Catalyst to Oil Ratio on Riser Temperature 64
3.7 Main Fractionator and Wet Gas Compressor 65
3.8 Regenerator 66
3.9 Regenerator Combustion Air Blower 67
3.10 Regenerator Lift Air Blower 68
3.11 Open-Loop Response of Regenerator Temperature
for 2% Positive Step Change in Coke Formation
Constant 69
3.12 Open-Loop Response of Reactor Temperature

Constant 70
3.13 Open-Loop Response of Carbon on Spent Catalyst
Constant 71
3.14 Open-Loop Response of Carbon on Regenerated
Catalyst for 2% Positive Step Change in Coke
3.15 Open-Loop Response of Flue Gas Oxygen
Concentration for 2% Positive Step Change
in Coke Formation Constant 73

for 2% Negative Step Change in Coke Formation
Constant 74
ix
Constant 75
Constant 76
Catalyst for 2% Negative Step Change in Coke

Concentration for 2% Negative Step Change
in Coke Formation Constant 78

for 20°F Positive Step Change in Fresh
Feed Temperature 79
Feed Temperature 80
Feed Temperature 81
Catalyst for 20°F Positive Step Change
in Fresh Feed Temperature 82

Concentration for 20°F Positive Step Change
for 20°F Negative Step Change in Fresh Feed
Temperature M
Temperature 85
Temperature 86

Catalyst for 20°F Negative Step Change in
X
Concentration for 20°F Negative Step Change
3.31 Open-Loop Response of Reactor Temperature for

2% Positive Step Change in Feed Rate 89
3.32 Open-Loop Response of Regenerator Temperature for
2% Positive Step Change in Feed Rate 90
in Feed Rate 91

2% Negative Step Change in Feed Rate 92
for 2% Negative Step Change in Feed Rate 93
in Feed Rate 94

Concentration for 30°F Positive Ramp in Ambient
Air Temperature 95

for 30°F Positive Ramp in Ambient Air Temperature 96
30°F Positive Ramp in Ambient Air Temperature 97
7.1 Unconstrained Control Response of Reactor

temperature for Disturbance 1 165
7.2 Unconstrained Control Response of Regenerator

Temperature for Disturbance 1 166
7.3 Unconstrained Control Response of Flue Gas

Oxygen Concentration for Disturbance 1 167




xi
7.10 Feed Rate Response with Constraint Control
for Disturbance 1 174
7.11 Regenerator Air Flow Rate Response with
Constraint Control for Disturbance 1 175
7.12 Firebox Temperature Response with Constraint
Control for Disturbance 1 176
7.13 Constraint Control Response of Reactor
7.14 Constraint Control Response of Regenerator
7.15 Constraint Control Response of Flue Gas
7.24 Plot of Gasoline Yield versus Gas Oil Conversion 188
7.25 Effect of Catalyst Coke Concentration on
Gasoline Yield 189
xii
D.1 Dynamic Response of Regenerator Temperature
for 0.05 Step Change in Fresh Feed Coke
Formation Factor 215
D.2 Dynamic Response of Reactor Temperature
for 0.05 Step Change in· Fresh Feed Coke
D.3 Dynamic Response of Carbon Concentration
on Spent Catalyst for 0.05 Step Change
in Fresh Feed Coke Formation Factor 217
D.4 Dynamic Response of Carbon Concentration
on Regenerated Catalyst for 0.05 Step Change
in Fresh Feed Coke Formation Factor 218
D.S Dynamic Response of Flue Gas Oxygen
Concentration for 0.05 Step Change in
Fresh Feed Coke Formation Factor 219
D.6 Dynamic Response of Regenerator Pressure
xiii
CHAPTER 1
INTRODUCTION
The Fluid Catalytic Cracking Unit (FCCU) receives

multiple feeds consisting of high boiling components from
several other refinery units and cracks these streams into
lighter and more valuable components. After further

processing, the FCCU product streams are blended with streams
from other refinery units to produce a number of products,
e.g., various grades of gasoline. Due to its large throughput
and ability to produce gasoline which is more valuable than
the feed, economic operation of FCCU is very important to the
overall economic performance of a refinery. Therefore,
considerable incentive exists to improve the economic
performance of FCCUs through the application of advanced
control and optimization strategies.
1.1. Nature of the FCCU Process
Generally the FCC process can be divided into three major
sections:
1. Reactor-Riser,
2. Regenerator,
3. Main Fractionator.
In a reactor riser, the feed is injected into a hot
stream of regenerated catalyst. The reaction occurs over a
short contact time in a riser before the catalyst and reacted
1
products are separated in the reactor. The catalyst residence
time in the riser tube is of the order of 10 seconds. The
short contact time minimizes the excessive deactivation of
catalyst and gives better yields of more valuable products
1 ike gasoline. The degree of gas oil conversion and coke
yield in the riser are a complex function of the reactor

temperature, catalyst-to-gas oil ratio and the condition
(activity, carbon concentration, metals content) of the
regenerated catalyst. After separation, the spent catalyst is
regenerated by burning off the carbon before circulating back
to the feed injector. The temperature in the reactor and the
regenerator will remain stable when the heat balance between
the two vessels is satisfied. This will occur when the heat
removed in the reactor to vaporize and crack the feed balances
the heat released by the combustion of coke in the
regenerator. Similarly, the carbon balance between the two
vessels should be satisfied, i.e., the carbon generated in the
reactor riser must be burned in the regenerator to keep the
carbon concentration on the catalyst as low as possible.
Thus, the effect of operating conditions on conversion and
yields in the reactor can be grouped into two simple terms.

One is the reactor temperature, the second can be called
"catalyst severity," which depends on the catalyst-to-oil
ratio, catalyst contact time an& catalyst activity. The unit
operator can measure the reactor temperature, but he has no
direct measure of the vital catalyst severity.
2
1.2. Objectives of An Ideal
FCCU Control System
Fluidized catalytic Cracker is one of the most difficult

processes in a petroleum refinery to control. It poses a
multivariable, strongly interacting and highly nonlinear

control problem. In addition, multiple hard equipment
constraints and operating constraints are present. With its
internal feedback loop created by the circulating catalyst and

its complex dynamic responses, the FCC unit is a truly
fascinating process to analyze and control. Its challenge to
the control engineer stems from these characteristics and from
the fact that the economic benefits are maximized when the
process is operated closer to constraint boundaries where
responses are more nonlinear. Economic objectives on
manipulated variables, such as the need to maximize either
charge rate or cracking severity, or both, compete with

operating constraints and further complicate the control
problem. The cat cracker must accept swings in feed stock
quality resulting from the variety of crude oils that today's
refineries process.
The objectives of FCCU control systems are:
1. Safety-Reverse catalyst flow protection,
2. Heat Balance-Maintain reactor and regenerator
temperature at their desired values within limits,
3. Carbon Balance-Catalyst regeneration to improve yields,
4. Conversion-Achieve most profitable level depending on
3
economics and unit constraints,
5. Interactions-Overcome process variable interactions to

achieve smooth control,
6. Constraints-Operate up to the physical limits
restricting further profitability.
1.3. Scope of Work
The first objective of this work is the development of a
dynamic simulator for Model IV Fluidized Catalytic Cracker.
The mathematical models presented by Georgakis et al. (1990)
will be used for this purpose. Their riser model includes

only the coke yield model for the reactor riser and does not
predict the composition of products. Therefore, a yield model
which will predict the yield of gasoline will be added in the
above model.
Once the dynamic simulator is ready, the second step will
be to study unconstrained control using nonlinear process
model based control strategy. This control strategy involves

use of approximate but phenomenological process models.
Therefore, these models will be developed for the reactor and
the regenerator. The unconstrained controllers will be tested
for setpoint changes and disturbances.
Once unconstrained control results are obtained, the
third step will be to develop a constraint controller for
constraints on feed pre-heater, air flow rate to the
regenerator and the catalyst circulation rate. The feed rate
4
to the system will be used as a manipulated variable. The
constraint controllers will calculate the maximum feed rate
which the system can process without hitting any of the
constraints. Therefore, constraint controller models will be
developed relating the feed rate to the various process
variables. These models will be used for constraint control.
The fourth part of this project is to do optimization
studies. The steady-state controller models will be used for
the optimization. In addition, steady-state models predicting
the gas oil conversion and yield of gasoline will be
developed. Initially the optimization will be done off-line
and parametric studies will be done. After successful off-
line optimization, the optimizer will be installed on-line

along with the controllers developed.
5
CHAPTER 2
LITERATURE REVIEW
This work involves development of a dynamic simulator of

fluidized catalytic cracker and the study of its control. So
the literature available can be divided into two main

categories:
i. Fluidized catalytic cracker process and its control,
ii. Multi-variable process control.
2.1. Fluidized Catalytic Cracking

Process and its Control
Literature related to fluid catalytic cracking process

can be separated into three sections:
i. Catalytic cracking and coke formation
ii. Regeneration of catalyst
iii. Dynamic simulation and control of FCC unit
2.1.1. Catalytic Cracking and Coke Formation
The catalytic cracking of petroleum gas oil results in a
broad spectrum of products ranging from hydrogen and methane
to heavy polymeric material adhering to the catalyst as coke.
This range of products is normally separated into marketable
fractions such as liquid petroleum gas, gasoline, light fuel
oil, etc. (Weekman et al. 1970).

weekman et al. (1970) reduced the broad spectrum of
catalytic cracking charge stocks and products into three
6
component system, namely, the original charge material, the
gasoline boiling fraction (C5 410°F) and finally the
remaining c4 •s (dry gas and coke). This simplified reaction
system is shown below:
P, .... a, P2 + a2 P3.
p2 .... P3.
Here, P1 represents the gas oil charged, while P2 represents
the c 5 - 410°F gasoline fraction and P3 the butane, dry gas and
coke. The a 1 and a 2 coefficients represent the mass of P2 and
P3 produced per mass of P1 converted, respectively.
Weekman (1968) proposed that the catalyst activity
changes with time, and the rate of decay of catalyst is first-
order in nature. Due to this, the rate of gas oil cracking is
a function of both time of exposure to oil and the fraction of
unconverted reactant. Weekman (1968) proposed that the
conversion of gas oil due to catalytic cracking in isothermal
fixed, moving, or fluidized bed reactors may be represented by
a pseudo-second-order reaction coupled with a first-order
decay of the catalyst activity.
Weekman et al. (1970) and Paraskos et al. (1976) found
that the activation energies for the gas oil cracking reaction
and the gasoline cracking reaction are different. Weekman
found that the gasoline cracking activation energy is almost
twice that of the gas oil cracking. Thus, as temperature is
lowered, the gasoline cracking rate declines more than the gas
oil cracking rate, and higher selectivity results.
7
During cracking, carbonaceous material, i.e., coke
deposits on the catalyst surface. Coke essentially contains
carbon and hydrogen. Voohries (1945) found that the
concentration of carbon on the catalyst surface can be
represented by following equation:
( 2 .1)
where
tc = contact time between catalyst and gas oil.

Pohlenz (1963) reported that value of n is approximately 0.5
and Kc depends on the type of catalyst, type of feed and

temperature of cracking. Jacob et al. (1976) reported that
value of n is close to 0.2 and Kc is a function of feed while
n is a function of catalyst type.
According to Kurihara (1967), there are mainly two

mechanisms for depositing coke: as catalytic carbon and as
additive carbon. The catalytic carbon is produced in the
cracking reaction, while the additive carbon is present in
most heavy gas oils and deposits without catalytic reaction.
By adopting Voohries (1945) catalytic carbon formation theory,

he proposed that the catalytic carbon formation is inversely
proportional to catalytic carbon content and inversely
proportional to some power of carbon content on the
regenerated catalyst.
Recently, Georgakis et al. (1990) presented a riser model
which can predict the amount of coke deposited on the
catalyst. The yield of coke deposited on the catalyst in the
8
riser is assumed to be effected only by the following: weight
hourly space velocity in the riser, concentration of carbon on
the regenerated catalyst, catalyst residence time in the riser
and the coking characteristics of various feeds. The effect
of riser temperature was not modeled. Both these models are
presented in detail in the next chapter.
2.1.2. Regeneration of Catalyst
The coke deposited on cracking catalyst is primarily
composed of carbonaceous deposits with hydrogen content
varying from 5 to 10 wt% (Weisz et al., 1966). Regeneration
of catalyst involves burning of coke with air to form carbon
monoxide, carbon dioxide and water. These exothermic
reactions take place in fluidized catalyst bed in the
regenerator. Weisz et al. (1963) proposed the kinetics of
coke combustion from a cracking catalyst. According to them,
coke combustion is a first-order reaction with respect to
both, concentration of coke on catalyst in the regenerator and
oxygen partial pressure in the regenerator. Thus,
I=kCrcY•' (2.2)
where
r = rate of carbon burning,
k = rate constant for coke combustion,
en = weight percentage of coke on catalyst in
regenerator,
Ym = oxygen partial pressure in regenerator.
9
According to Gates et al. (1979), rate of carbon burn off
is a function of oil feed stock from which it is formed. For
example, if the oil contains a very small amount of
organometallic compounds, it results in metal deposit on the
catalyst and these metal deposits may accelerate coke
combustion. Thus the rate constant for carbon burning may
change from case to case depending upon the feed stock from
which coke is formed.
Ford et al. (1976) presented a detailed regenerator model

which can predict the dynamic response of the catalytic
cracking regenerator to changes in operation. Based upon
fundamental kinetics, dynamic material and energy balances and

catalytic cracking operating principles, this model can be
used to evaluate current operations, transitions from one
operating region to another and new designs and control
systems. The model assumes that conversion of hydrogen
present in the coke on the spent catalyst to water is complete
according to following reaction:
I • 4 H + 0 2 .... 2 H 20 •
Carbon in the coke reacts with oxygen to produce carbon
monoxide and carbon dioxide in the following reactions.
II • 2 C + 02 .... 2 CO,
III • C + 02 .... C02 •

carbon monoxide reacts to produce carbon dioxide
according to:
IV. 2 CO + 0 2 .... 2 C02 •
10
Initial gas and catalyst separation occurs in the disengaging
zone above the bed. As long as catalyst is present,
combustion of carbon occurs according to reactions II and III.
In the region above the disengaging zone, only the combustion
of carbon monoxide occurs. The fluidized bed is assumed to
consists of two phases: a bubble phase of gaseous reactants
and products moving up the bed in plug flow and a perfectly
mixed dense phase containing gas and solid catalyst. Mass
transfer occurs between the two phases as gas moves up the
bed. But at typical regenerator temperatures, reaction rates
are controlling rather than the rate of mass transfer between
the two phases. Hence, the gas velocity has no effect on rate
of carbon burning. Also, they observed that there are no
significant differences in the regeneration characteristics of
zeolite catalysts and silica-alumina
0
catalysts. Since this
regenerator model is used for the development of dynamic
simulator in this work, the kinetic expressions for the above
reactions and regenerator mass and energy balances equations
are given in the next chapter.
2.1.3. Dynamic Simulation and

Control of FCC Unit
2.1.3.1. Dynamic Simulation of FCC Unit

Kurihara (1967) presented a dynamic model for a FCC unit
which he used for control and optimization studies. During
developing a dynamic model, he assumed that:
1. A fluidized bed in the regenerator is perfectly mixed,
11
2. Vaporized feed stock passes through the reactor bed in a
plug flow manner with negligible time lag,
3. Constant pressure is maintained in both vessels,
4. Activation energies, heats of reaction and heat of
vaporization of the feed are all constant,
5. Carbon buildup on catalyst is inversely proportional to

carbon on regenerated catalyst.
Kurihara developed dynamic models of reactor and regenerator

involving carbon balances, heat balances and reaction rate
expressions.
Lee et al. (1985) presented a mathematical model of a

modern catalytic cracking plant using a zeolite catalyst.
They used the kinetic expressions published in the literature.
Some of the features in this model are:

1. Riser is assumed to be an adiabatic plug flow reactor,
2. Model neglects the changes with time during the passage
of catalyst through the riser (Quasi-steady state),
3. Regenerator is assumed to be perfectly mixed,

4. The pressure is assumed to be constant in both riser and
regenerator.
The dynamic simulator developed was used for testing different
PI control configurations.
Recently, Georgakis et al. (1990) presented a dynamic
simulation model for the reactor/regenerator section of a
Model IV FCC unit. This model is sufficiently complex to
capture the major dynamic effects in a Model IV
12
reactor/regenerator, including nonlinearities and
interactions, as well as equipment and operating constraints.
Some of the salient features of this dynamic model are,
1. Reactor is assumed to be a perfectly mixed tank,

2. Riser energy balance also assumes stirred tank dynamics,
3. Fluidized catalyst bed in the regenerator is assumed to

be perfectly mixed,
4. Pressures in the reactor, main fractionator, regenerator

are modeled.
Since this work has made extensive use of this dynamic
model, it is described in detail in the next chapter.
2.1.3.2. Control of FCC Unit

Fluid catalytic cracker is one of the most difficult
processes in a petroleum refinery to control due to presence
of nonlinearities and coupling. FCC process cracks millions
of barrels of oil daily worldwide. Because of the large
volume of oil handled, even a small increase in efficiency
pays important dividends in terms of energy savings and
profits. Therefore, tighter and more efficient control of FCC

unit is an important objective. FCC unit is frequently
subjected to disturbances like changing coking characteristics
of feed stock, feed rate changes, feed temperature changes

etc. so the control should be effective to operate the
process near the constraint boundary and to reduce the
disturbance effects at the same time.
13
A simple method of controlling the process is frequently
described in the literature (Pohlenz, 1963: Venuto et
al.,l979) is the "conventional control" scheme. This control

configuration is depicted in Figure 2 .1. In this scheme,
catalyst circulation rate is manipulated to control reactor
temperature while air flow rate to the regenerator is
manipulated to control flue gas oxygen content. Proportional-
Integral controllers are the usual choice for this
arrangement. In practice oxygen concentration in flue gas is
not measured directly, but temperature difference between
regenerator dense bed and regenerator stack gas is used as an
indicator of flue gas oxygen content. High regenerator stack
gas temperature is an indication of afterburning caused by a
high level of oxygen in the flue gas. Although this control

scheme has been in operation on many units for many years, it
has following disadvantages. (Kurihara, 1967: Gould et al.,
1970).
1. This control scheme is sluggish being unable to bring the
process quickly back to desired operating conditions,
2. It exhibits large swings in regenerator temperature which
is unacceptable if the unit is operated near the maximum
regenerator temperature.
In 1967, Kurihara (Kurihara, 1967: Gould et al., 1970)
developed a new FCC control scheme. This scheme is depicted
in Figure 2.2. In the Kurihara control scheme, air flow rate
to the regenerator is manipulated to control the regenerator
14
temperature while catalyst circulation rate is manipulated to
control the flue gas oxygen content. It can be seen that in
this scheme the control emphasis is on regenerator and the
reactor temperature is not controlled directly. ·It is the

most critical variable the operators watch and control
closely, because it strongly affects the product yields and
the performance of the downstream distillation column.
Therefore, the obstacle of implementing this scheme is to
convince the operators of the merit of this scheme. It is not

known if this scheme is widely used or not commercially.
Based on Kurihara's scheme, Lee et al. (1973) developed

a new control scheme in which two modifications from
Kurihara's scheme were incorporated. First, cascade control
was added to adjust the regenerator temperature setpoint as

the reactor temperature varies. This gave the operator the
flexibility of changing the distribution of temperature
excursions between the two vessels. Secondly, the flue gas
temperature instead of the difference between the flue gas and
the dense bed which is an indicator of oxygen concentration in

flue gas, was used to manipulate the catalyst circulation
rate. This was designed to eliminate the difficulties
associated with the phase shift between the flue gas and dense
bed temperatures. The advantage of this modified scheme over
Kurihara's control scheme is that it better accommodates the
operator's concerns and passes fewer disturbances to
downstream units.
15
Bromley and Ward (1980) proposed a "Structural Analysis
Control" scheme which offers loop uncoupling, feedforward
disturbance compensation and internal-variable compensation
based on the state-vector model. This approach uses
proportional-mode approximations to the model-based
controllers, along with PI feedback controllers. The control
logic is primarily feedback control and the method provides
auxiliary model-based compensation activated by the controlled
outputs, the manipulated inputs, the internal variables, and

the measured disturbances. Each model-based controller
utilizes the functional dependence of a single measurement and
manipulation on a single controlled-variable derivative.
While less sophisticated than most multivariable approaches,
structural analysis control is simple and effective way to use

model and measurement information while still retaining the
proven multiloop feedback structure. Schuldt et al. (1971)
presented "Quadratic Performance Synthesis" scheme which is
derived from optimal control theory. Cutler et al. (1992)
recently presented the experiences applying dynamic matrix
control (DMC) on a commercial Model IV FCC unit. In order to
stabilize the carbon on catalyst, they used measurements of
the regenerator bed temperature and the flue gas oxygen

concentration to infer the instantaneous value of the carbon
on the regenerated catalyst, using a simple regenerator bed
model. Recently, Grosdidier et al. (1993) presented a paper
describing the implementation of IDCOM-M control strategy on
16
a fluid catalytic cracking unit. They have discussed the
challenge which FCC unit offers to control engineers by virtue

of its complex process dynamics and because of the conflict
between its operating constraints and its economic objectives.
They have also discussed the dynamic process models identified

from open-loop test data.
Prett et al. (1980) presented a paper on optimization and
constrained multivariable control of a catalytic cracking
unit. They developed a steady state model of the process to
generate partial derivatives relating all the pertinent
variables. These derivatives were used in a linear program to
find the economic optimum for the operation. They have

reported enhancing the dynamic matrix control algorithm to
recognize the constraints. The algorithm recognizes the

present and future constraints and calculates the future
values of the controlled variables rather than just the
feedback or present values. Hence, future violations of the
controlled variable constraints can be predicted in advance
and dealt with in an anticipatory mode.
2.2. Multivariable Process Control
Literature in the area of multi-variable process control
is very extensive. Various advanced control strategies have
been proposed in the past for control of both single input,
single output (SISO) processes as well as multi-input,
multi-output (MIMO) processes. These advanced control
17
strategies can be broadly classified into two categories,
strategies based on linear models and strategies based on

nonlinear models.
2.2.1. Linear Model Based Control

Linear model based control strategies use linear,
empirical process models. Dynamic Matrix Control, DMC
(CUtler, 1983), Internal Model Control, IMC (Garcia and
Morari, 1982) and IDCOM (Setpoint, 1979) belong to this class

of control strategies.
DMC uses a step response linear model. DMC is a time
horizon controller that use hypothetical disturbance level at

each step to remove any processjmodel mismatch and uses this
calculated disturbance level for the prediction horizon. The
user tunes this controller by selecting the prediction

horizon, and the weighing factors for the input moves. DMC is
widely used in a number of industrial applications.

Internal Model Control, IMC uses transfer function models
of the process. It calculates the control action necessary to
return the process to the desired setpoint. A filter is
applied to the feedback signal in order to improve the
robustness. The IMC structure feeds back the process/model
mismatch to make adjustments to the setpoint in order to
remove the offset.

IDCOM uses a weighted time series model of the process
which is basically a convolution model of the process. It
18
assumes that the prediction horizon is equal to the control
horizon. It is similar to DMC except it uses an impulse
response model instead of step response model.
A major advantage of linear model based control

strategies over conventional PID control is that these methods
take into account the multi-variable nature of the process,
and as a result consider coupling in the process. Unlike
conventional methods, these algorithms can deal with process
constraints, such as bounds on state and manipulated variables

in an explicit manner. These controllers are thought to be
simpler to tune for robustness to modeling error. The
limitations of linear model based control are (1) they
consider only linear picture of the process, (2) the models
used are only locally valid, (3) extensive plant testing is
required to identify the proper linear models, and (4) the
models are not useful for optimization. As operating
conditions change, linear model based controller must be de-
tuned in order to get operational reliability. Also, process
re-identification of the model parameters.
2.2.2. Nonlinear Model Based Control

Due to limitations of linear model control mentioned
above, nonlinear model based control strategies like Generic
Model control, GMC (Lee and Sullivan, 1988), Nonlinear
Predictive Model Control, NLPMC (Parrish et al., 1988) have
been proposed.
19
Generic Model Control uses a nonlinear model of the
process which includes phenomenological aspects of the

process. To understand GMC, let us consider a single input,
single output (SISO) process described by the following model:
~~ = f {y, u, d, k) , (2.3)
where y is the output variable, u is the manipulated variable,

d is a vector of measured disturbance, and k is a vector of
adjustable parameters. Assuming that y has a value of y 0 , and
it is desired to move the process from y 0 to Y.p in some
interval ~, then equation (2.3) can be approximated using the
forward difference approximation of the first derivative:
Ysp - Yo = f( Yo , u, ~,
d .c.r,.) • (2.4)
"'
Then equation (2.4) can be solved directly to determine u, the
control action, if we know ~ and k. The time interval ~ is
a tuning parameter. If ~ is small, rapid response will be
obtained and if ~ is large, sluggish response will be
obtained.
Since the control model used in equation (2.4) is not

exact, use of this control law will result in steady-state
offset. To eliminate this offset, Lee and Sullivan added an
integral term, resulting in the GMC control law given by
equation (2.5).
20
(2.5)
K1 is equal to 1/~ in equation (2.4). Thus this equation can
be solved to calculate the control action. For a multiple
input, multiple output (MIMO) control configuration, the

control laws can be written for all the loops and the
resulting set of equations can be solved simultaneously to get
the desired control action.

Economou et al. (1986) proposed Nonlinear Internal Model
Control which uses nonlinear model instead of linear model
used by Internal Model Control, IMC (Garcia and Morari, 1982).

However, it is implicit and requires an iterative integration
of the model over a time horizon to calculate the control

action.
This work involves use of nonlinear process models with
Generic Model Control as a control law. Nonlinear PMBC has

been successfully applied to processes, such as coal
gasification (Pandit et al., 1989), pH (Choi et al., 1986;

Williams et al., 1990; Mahuli et al., 1992), commercial
distillation (Riggs et al., 1989), wastewater neutralization
(Rhinehart, 1989), supercritical fluid extraction
(Ramachandran et al., 1992).
21
TO f"RACTIONATOR
f"LUE GAS
PIO
REACTOR
SET~-- PID
rSET
I I
~-
r---- ~- ___ 1
I
AIR
~--- OIL FEED
Figure 2.1 Schematic of Conventional Control Scheme
22
TO fRACTIONATOR
fLUE GA S
REACTOR
REGENERATOR
s::r o--jrrg j-
I
D..
AIR
Figure 2.2 Schematic of Kurihara Control Scheme
23
CHAPTER 3
PROCESS DESCRIPTION AND DYNAMIC SIMULATION
3.1. Process Description

The process flow diagram of FCC unit is given in Figure
3.1. The fresh feed is pre-heated in a furnace and is mixed
with hot slurry recycle coming from the bottom of the main
fractionator. The mixed feed is injected into the reactor
riser, where it mixes with hot regenerated catalyst coming
from the regenerator and completely vaporizes. Cracking is an
endothermic reaction and the hot catalyst provides the
sensible heat, heat of vaporization and heat of reaction.
During the cracking reactions, a carbonaceous material (coke)
is formed, which gets deposited on the surface of the
catalyst. Coke consists of carbon and hydrogen, proportion of
which depends on the coking characteristics of the feed stock.
Since coke poisons the catalyst, continuous regeneration of
the catalyst is necessary. Separation of catalyst and the
product gases occurs in the disengaging zone of the reactor.
Reactor product gas is passed to the main fractionator for
heat recovery and separation into various products. Spent
catalyst falls in the catalyst bed in the reactor from where
it goes to the regenerator through the spent catalyst U-bend.
Air is injected into the bottom of the regenerator lift
pipe to assist the circulation of catalyst. Catalyst in the
regenerator is fluidized with air flow provided by the lift
24
and combustion air blowers. Carbon and hydrogen on the
catalyst react with oxygen to produce carbon monoxide, carbon
dioxide and water. Most of the reactions occur in the
fluidized catalyst bed, but some reactions do occur in the
disengaging zone above the bed. Gas travels up the

regenerator into the cyclones where entrained catalyst is
removed. Regenerated catalyst flows over a weir into the
I
regenerator standpipe. The level of catalyst in the standpipe
provides the driving force for catalyst flow to the reactor
riser. Bead in the catalyst circulation lines and vessels,

reactor/regenerator differential pressure and lift air flow
rate (affecting catalyst bulk density in the regenerator lift
pipe) can all be manipulated to alter the catalyst circulation

rate, an important manipulated variable. The models equations
for various units in a FCC process are selected from
literature. They are described below. The nomenclature and
values of model constants are provided in the Appendix A.
3.2. Feed Preheat System
The schematic of feed preheat system is given in Figure
3.2. This model was taken from Georgakis et al. (1990). The
fresh feed (Gas Oil) is assumed to enter the preheat furnace
at a specified temperature, T1 • This temperature is a
measured disturbance variable in the simulation that can be
preprogrammed as a function of time for disturbance rejection
testing of control systems.
25
The furnace firebox temperature and outlet temperature
are modeled with the following assumed dynamic energy
balances.
dT3 =
dt (3.1)
where
( T3 - Tl ) - ( T3 - T2 )
I
ln( T3- Tl) (3.2)
T3 -T2
(3.3)
F.s = .rs
l:'set
1 (3.4)
( 3. 5)
(3.6)
3.3. Reactor-Riser Model
Figure 3.3 is a schematic of a reactor-riser section.
While developing dynamic simulator for a FCC unit, two
different reactor riser models given in the literature were

considered. Both the reactor-riser models are described
below. For distinguishing between the two models, they are
identified by following names:
26
i. Amoco Model (Georgakis et al., 1990),
ii. Lee Model (Lee et al., 1985).
3.3.1. Amoco Model

ln this model, the reactor riser is simulated as a dilute
phase solid transport line in which hot catalyst from the
regenerator mixes with feed oil from the preheat system. All
of the cracking reactions are assumed to occur only in the
reactor riser. The riser yield model predicts only the amount
of coke deposited on the catalyst and the amount of wet gas
(C 6 's and lighter) produced. A constant coke composition is
assumed. The yield model simulates the feed composition
changes through factors which alter the coke producing
characteristics of various feeds. the composition of wet gas
is not predicted. Pressures in the main fractionator and
reactor are modeled.
3.3.1.1. Coke and Wet Gas Yield Models

The yield of coke deposited on the catalyst in the riser
is assumed to be effected only by the weight hourly space
velocity in the riser (WBSV), the concentration of carbon on
regenerated catalyst ( crgc) ' catalyst residence time in the

riser (~r) and the coking characteristics of various feeds.
The effect of riser temperature on coke yield is not modeled.
The coke production is given by
27
77. 03 ( F3 + F4 ) Fb -r;0 • 69
I (3.7)
100 WHSV
where
't r = (3.8)
3 6 0 0 ( F 3 + F~ )
WHSV = • I (3.9)
(3.10)
= 1 normal gas oil

> 1 heavier than normal gas oil
< 1 lighter than normal gas oil
> o.
The yield of wet gas is assumed to be affected only by the
riser temperature and is given by
(3.11)
3.3.1.2. Reactor Mass Balance

Coke and catalyst balances are written for the reactor.
Stirred tank dynamics are assumed. The coke balance includes
the carbon entering the reactor on regenerated catalyst and is
given by
28
dCsc =
(3.12)
dt
The catalyst balance is give·n by
(3.13)
3.3.1.3. Reactor Riser Energy Balance

The riser energy balance also assumes stirred tank
dynamics, with negligible heat loss to the environment. The
heat of cracking is assumed to be simply proportional to riser
temperature. The energy balance is written as:
(3.14)
where
(3.15)
Orout = Ocatout + Oslurry + Ocracking + Qff' (3.16)
(3.17)
(3.18)
(3.19)
(3.20)
(3.21)
29
A constant temperature drop across the reactor stripper is
assumed.
(3.22)
3.3.1.4. Reactor Riser Pressure Balance

The pressure at the bottom of the reactor riser is needed
in the force balance on the regenerated catalyst U-bend. It
is given by
= p + Prishris (3.23)
f, 144 I
where
(3.24)
= F3 + F4 + Frgc •
(3.25)
Pv Ppart
The holdup of catalyst in the riser is required in the

calculation of the catalyst residence time in the riser. It
can be calculated as follows:
= FrgcArishris
(3.26)
vris
3.3.1.5. Reactor and Main Fractionator Pressure Balances

The main fractionator is modeled as a single large volume
that includes the volume of piping and intermediate storage
30
vessels between the main fractionator and the wet gas
compressor. It is modeled with assumed dynamics given by
dP5 = 0.833 (F~-Fv. -Fv.u +Fv.13 ) , (3.27)

dt -~ 11
where
and Fv11 and Fv13 are given in Section 3. 4. A constant pressure
drop between the main fractionator and the reactor is assumed,

giving the reactor pressure as
(3.29)
3.3.2. Lee Model

The Lee model for the catalytic cracker is taken from Lee
et al. ( 1985) • The assumptions involved in this model and
model equations are described below. The details of this
riser model such as derivation of model equations, estimation
of model parameters can be obtained from E. Y. Lee ( 1984).
This model treats the riser as an adiabatic plug flow reactor.
It is a quasi-steady-state model in that it neglects changes
with time during passage of feed through the riser. Since the
residence time of feed through the riser is less than 10
seconds, and the characteristics time for changes in the
regenerator is of the order of hours, this is a reasonable
approximation. The model consists of the following material
and energy balances:
31
Material balance on gas oil:
(3.30)
where
(3.31)
Material balance on gasoline:
Energy balance: (assuming adiabatic operation)
d6 = A.llHfFo dy f
(3.33)
dz T(O) ( F 8 CPc + A.F0 C.Po + (1- .A) F 0 Cp.) dz '
where
e (z) = T(z) -T(O)

T(O) I
(3.34)
= FB c.Pc Treg +A. Fo c.Po To+ ( 1- A.) Fo cp, To

T(O) (3.35)
FBCPc +A.FOCPo + (1-A.) FoCp, I
Carbon buildup on the catalyst is assumed to be affected
by catalyst residence time in the riser (tc), riser
32
temperature and carbon on regenerated catalyst ( Crgc) and is
given by following equation:
(3.36)
where
T(1) = T(O) (1+8(1)) I (3.37)
The nomenclature of variables and values of various

parameters used in above equations in Lee Model is given
below. These values were taken from E.Y. Lee (1984).
ccat = weight fraction of carbon deposited on catalyst,
Crgc = weight percent of coke on regenerated catalyst,
COR = catalyst to oil ratio, lb/lb,
Ef = activation energy of gas oil cracking,

43700 BTU/lbmol,
E = activation energy of gasoline cracking,
9
48500 BTU/lbmol,
Ecf = activation energy of coke formation in riser,
18000 BTU/lbmol,
a = catalyst activity decay constant, 0.1/sec,
1 = weight fraction of gas oil in feed stream to riser,

8 - dimensionless temperature variable,
'
0
= catalyst decay function,
AHf = heat of endothermic reaction of gas oil cracking,
218 BTU/lb,
33
= reaction rate constant, 9. 6 x 10 5 s - 1 ,
= reaction rate constant, 7.2 x 10 5 s- 1 ,
= reaction rate constant, 4. 224 x 10 5 s- 1 ,

= catalytic carbon formation rate constant, 0.019 s -1 ,
n = empirical exponent, 0.4,
C~ ~ specific heat of catalyst, 0.24 BTU/lb R, 0
C~ = specific heat of oil, 0.75 BTU/lb0 R,

C~ = specific heat of dispersing steam, 0.45 BTU/lb0 R,
R = universal gas constant, 1.9872 BTU/lbmol 0 R,
tc = catalyst residence time in riser, seconds,
0
T0 = temperature of gas oil feed, R,
Yt = weight fraction of gas oil in vapor in riser,
y9 = weight fraction of gasoline in vapor in riser.
The major differences in these two models are as follows:
1. The Amoco model assumes stirred tank dynamics for the
reactor riser energy balance while the Lee model assumes
plug flow behavior for the mass and energy balances,
2. The Amoco model predicts only the amount of coke
deposited on the catalyst and the amount of wet gas
produced. It does not predict the yield of gasoline.
The Lee model predicts both the amount of coke deposited
on the catalyst as well as the yield of gasoline in the
reactor riser. The prediction of gasoline yield is
necessary for economic optimization of FCC unit,
3. The Amoco model neglects the effect of reactor riser
temperature on the yield of coke while the Lee model
34
includes the effect of reactor riser temperature on the
production of coke.
In this work, both reactor-riser models were studied. To
make steady-state comparison between two models, the models
were first parameterized to match at the base case. The heat
of cracking reaction and carbon formation rate constant were
the parameters which were used for this purpose. The results
of steady-state comparisons are shown in Figures 3.4-3.6. It
was found that there was not a large difference in the two
models. It was decided to use Lee model for the reactor riser
since it can predict the yield of gasoline, which is important
to know for doing economic optimization of FCC unit. Also,
the Lee model assumes plug flow behavior for the reactor riser
which is close to reality as against perfectly mixed behavior
assumed by Amoco model. For modeling reactor and main
fractionator pressures, part of Amoco model discussed above is
used. After the Lee model was selected for the reactor riser,
the time delay associated with the passage of catalyst from
reactor riser to where it enters regenerator was added. This
time lag essentially corresponds to the movement of catalyst
in a catalyst bed outside riser tube and it was found out to
be 100 seconds in magnitude.
3.4. Wet Gas Compressor

Figure 3. 7 is a schematic of the wet gas compressor. The
wet gas compressor pressurizes gas to the operating pressure
35
of downstream separators in the vapor recovery unit. The
model equations for the wet gas compressor are taken from
Georgakis et al. ( 1990) • The wet gas compressor is modeled as
a single stage centrifugal compressor driven by a constant
speed electric motor. It is assumed that the compressor is

pumping against the constant pressure in the vapor recovery
unit (vru). The compressor performance equation relates
suction flow to polytropic head. A single surge point is
specified. Two bypass lines with valves are available for
implementation of anti-surge control; a vent line to the
atmosphere on the suction line of the compressor and the
bypass line between the compressor discharge and inlet lines.

However, the wet gas compressor is normally operated far from
surge and for the purpose of this simulation, anti-surge
controllers can be excluded.

The performance equation is given by
F sucn,wg = 11 , 6 0 0 + J1 . 3 6 6 X 10 8 - 0 . 1 0 57 H!. ,
"~
(3.39)
where
H wg = 182,922.1 (C~_;
0942
-1), (3.40)
Pvru
c = p;.
nf
(3.41)
The molar flow rate through the compressor is obtained from
the ideal gas law:
36
(520) Fsucn,wgp7
(379) (60) (590) (14.7) •
(3.42)
The pressure balance around the compressor is given by assumed

dynamics:
dP7
= (3.43)
dt
where
(3.44)
(3.45)
Certain valve flow characteristics are nonlinear and are given

by:
fPP ( x) = exp ( 2ln ( o .15 ( 1-x) ) ) , X> 0. 5
(3.46)
X::SO.S (3.47)
3.5. Regenerator
Spent catalyst from the reactor is transported through
the spent catalyst U-bend and enters the regenerator through
a lift pipe. Pressurized air (lift air) is injected into the
bottom of the lift pipe to assist in catalyst circulation.
Coke is burnt off in the regenerator by contacting the
catalyst with air in a fluidized bed. Combustion air,
delivered directly to the bottom of the regenerator by the
37
combustion air blower, and lift air both contribute to the
fluidization and combustion reactions.
The regenerator model is taken from Georga~is et al.

( 1990) • The model assumes that conversion of hydrogen present
in the coke on spent catalyst to water is complete according
to the following reaction:
I. 48 + 0 2 -. 28 2 0.
Carbon in the coke reacts with oxygen to produce carbon
monoxide and carbon dioxide in the following reactions.

II. 2C + 0 2 -. 2CO,
III. C + 0 2 _. C02 •
Carbon monoxide reacts to produce carbon dioxide according to:

IV. 2CO + 0 2 _. 2C02 •
Initial gas and catalyst separation occurs in the

disengaging zone above the bed. As long as catalyst is
present, combustion of carbon occurs according to reactions II
and III. In the region above the disengaging zone (referred

to as the dilute phase) , only the combustion of carbon
monoxide occurs. Regenerated catalyst flows over a weir into

the regenerator standpipe for transport back to the cracking
riser.
A detailed description of the derivation of the
regenerator mass and energy balances and the kinetic equations
for the reactions can be found in Ford et al. ( 1976). A
schematic of the regenerator is given in Figure 3.8.
38
3.5.1. Energy Balance
The fluidized catalyst bed in the regenerator is assumed
to be perfectly mixed. The energy balance for the regenerator
bed is given by following equations:
(3.48)
(3.49)
(3.50)
(3.51)
The amount of heat generated by hydrogen combustion is given

by following equation.
(3.52)
The quantity of heat produced by reactions I-III is given by
(3.53)
(3.54)
(3.55)
(3.56)
The amount of hydrogen burned off in the regenerator assuming
total conversion is given by

39
(3.57)
Catalyst is present above the bed due to carry over. The

temperature increases above the bed due to heat generated by
reactions I-III. The heat balance between the top of the bed
and the inlet of the cyclones is affected by the presence of
catalyst in this zone. The following equation gives the
temperature profile above the catalyst bed.
dTreg (z) =0 1 O~z~zbed (3.58)

dz
dTreg (z) = ( ~H dXco (z) +~H dXco2 (z) ) 1

1 dz 2 dz C.P(z)'
dz
• • • • • • • • • • zbed-<Z=SZT (3.59)
(3.60)
(3.61)
3.5.2. Carbon Balance

The following equations represent the catalyst and carbon
balance in the regenerator.
dCrgc (3.62)
dt
(3.63)
40
(3.64)
3.5.3. Standpipe Inventory Balance
Catalyst flows over a weir and into the regenerator

standpipe according to following equations. When the catalyst
level in the standpipe is greater than 17 ft (the standpipe
height is 20 ft) the catalyst flow over the weir is slowed.
(3.65)
(3.66)
(3.67)
(3.68)
3.5.4. Mass Balance

Mass balances on oxygen, carbon monoxide and carbon
dioxide are given below. The derivation of these balances can
be found in Ford et al. (1976). The concentrations of carbon
monoxide and carbon dioxide in the air entering the
regenerator are assumed to be zero. The amount of oxygen

available for reactions I-III described above is that
available in the air flow entering the regenerator minus the
41
oxygen consumed in the combustion of hydrogen. The rates of
reactions I and II are affected by the volume fraction of the
catalyst (Ps(z)) in the bed and the disengaging zone above the
bed. The change in Ps(z) as a function of vertical position
in the regenerator is described below.
dX~ (z)
dz
(3.69)
dXco(Z)
(3.70)
dz
dXOa ( z) dX00 ( z)
(3.71)
dz -o. 5 dz
(3.72)
(3.73)
9 88 34,000 ] . (3.74)
kl = 6.9547exp [ 1 . - (Treg(z) +459.6)
25 000
k
2
= 0.69148exp[15.06- (
'
Treg(z) +459 .6)
] · (3.75)
(3.76)
The output concentrations are converted to the appropriate
units as given below.
42
(3.77)
6
C = (lx10 ) 28Xco
co,sg 28Xco+44Xco.t +32X~ +22 .12
(3.78)
Three minute analyzer delay is then added for these

measurements.
3.5.5. Volume Fraction of Catalyst

Void fraction in the regenerator bed (ef) is a function
of superficial velocity and is given by empirical equations
given below. Volume fraction of catalyst ( Pa(z)) is given by
following equations. In the disengaging section above the
bed, catalyst volume fraction drops off vertically. It is
assumed that any entrained catalyst entering the cyclones is
removed and returned to the bed so that ( p8 <•)) is zero in the
cyclones.
= 0' (3.79)
(3.80)
dp 8 (z) = (3.82)
0 I
dz
43
p 8 (z) = 01 (3.83)
e = min[l max(e 1.904+0.363v8 -0.048~

8 I fief+ Z )].
bed
(3.84)
ef = o. 332 + o. 06 v •• (3.85)
3.5.6. Catalyst Entrainment

The mass flow rate of entrained catalyst leaving the bed
is described empirically by following equations.
(3.86)
Pc,dilute = -0 • 878 +0. 582 V 8 • (3.87)
Superficial velocity is given by
= ( Fsg +Fair ) ( 1 (3.88)

VB J1 ) 1
2 P'T·reg
where
- (520) p6
(3.89)
Pg- (379) (14.7) (Treg+459.6)
3.5.7. Pressure Balance

The pressure balance in the regenerator assumes ideal gas
behavior. The pressure at the bottom of the regenerator, Prgb'
is required in the calculation of air flow into the
regenerator from the air blowers.
44
1 [ R ( n d Treg + ( T + 4 59 6 ) d n 1
vreg,g dt reg • dt , (3.90)
(3.91)
(3.92)
P = P.6 +
w.reg
rgb 144Areg (3.93)
(3.94)
Stack gas flows from the regenerator through the stack gas
valve, V14 , as given by following equation.
(3.95)
3.5.8. Bed Height

The regenerator bed height is given empirically by
following equation.
z bed = min [ z eye , ( 2 •85 +0 •8 v s + - W..:r.;;;egL.-._..;,_p.=.:c,~di:.:l:.::u:.;.;te;;..A...;r;.;:eg£-Z~cy~c ) ( 1 )

AregPc,dense _ Pc,dilute
1
Pc,densB
(3.96)
where
Pc, dense = Ppart ( 1-ef) • (3.97)
45
3.5.9. Air Lift Calculations
Spent catalyst from the reactor enters the regenerator
through the lift pipe. Air from the lift air blower is
injected into the bottom of the lift pipe to assist the
circulation of catalyst. An increase in lift air flow rate
lowers the density of catalyst in the lift pipe which in turn
lowers the head, resulting in an increase in catalyst flow
through the spent catalyst U-bend.
Changes in density in the lift pipe occur with an assumed
time constant, l:fill' and are given by following equation.
dp lift; =( FBC
A +pairg-plift
)
--
1
1 (3.98)
dt Vcat;, lift; lp 't fill
where
(3.99)
Pairg = R ( Tsc+459 • 6) I
F9 (3.100)
vair,lift : _ _:;___I
AlpPairg
(3.101)
The pressure at the bottom of the lift pipe is the sum of the
regenerator pressure, P6 , and the heads in the lift pipe and
46
the regenerator bed above the top of the lift pipe.
(3.102)
3.6. Air Blowers

Air is supplied to the regenerator by two centrifugal
compressors. The bulk of regenerator air is supplied by the
combustion air blower, which supplies air directly to the
bottom of the regenerator. The lift air blower injects air
into the bottom of the lift pipe to assist the circulation of

spent catalyst from the reactor to the regenerator. At a
given catalyst circulation rate, excess lift air blower
capacity may be available. In this situation, air may be
bypassed from the lift air blower directly to the bottom of
the regenerator through the spill air line. The model

equations for the blowers are taken from Georgakis et al.
(1990).
3.6.1. Combustion Air Blower
The schematic of combustion air blower is given in Figure

3.9. The combustion air blower is a centrifugal compressor
driven by a constant speed electric motor. Throughput is
controlled by a throttling valve on the compressor inlet line.

An atmospheric vent line and valve are available on the
discharge line for use in an anti-surge control system. A

head-capacity performance equation is provided that relate
47
suction volume as a function of discharge pressure when the
suction is at normal atmospheric pressure. To obtain suction
flow rates at different suction pressures, it is assumed that
the performance curve at any suction pressure is the same as
the curve for one atmosphere at the same polytropic head. A
single surge point on the performance curve is specified.

The head-capacity performance equation is given by,
Fsucn,comb = 45, 000 +J 1. 581X10 9

-1.249 x10 6 Ptase I ( 3.103)
where
(3.104)
Suction and discharge pressures are given by following
equations:
(3.105)
(3.106)
where
(520) (29) PlF•ucn,cOIIlb (3.107)

F.6 = ~----~~~~~~~~~~~'
(14.7) (379) (60) (T.tm+460)
(3.108)
48
(3.109)
(3.110)
3.6.2. Lift Air Blower
Figure 3.10 is a schematic of the lift air blower. The

lift air blower is a single stage centrifugal compressor
driven by a variable speed steam turbine with a speed control
governor. A local controller manipulates a steam valve for
speed adjustment. An atmospheric vent line and valve are
available on the discharge line for use in an anti-surge
control system. A single performance equation is provided
that relates polytropic head divided by speed squared and
suction volume divided by speed. Suction flow for a given
discharge pressure is returned at a specified base speed and
then corrected for actual compressor speed. A surge line is
defined, relating surge flow rate to discharge pressure.
The performance equation is given by following equation:
Fbase = 8600 +J2. 582X 10 8 -1. 068X 10 5 Ptase,d, (3.111)
where
(3.112)
M= (3.113)
49
Lift air flow rate is obtained from:
(52 0) ( 29) PatmFsucn, lirt

Fa = ~----~------~~~~~---
( 14 . 7 ) ( 3 7 9 ) ( 6 0) ( TIJ tm + 4 6 0 ) I
(3.114)
where
- ( Ba)
F sucn, 11~t - F base S 1 (3.115)
b
Fsuxge,lUt = 5025 +112P3 • (3.116)
Flow rates associated with lift air blower system and spill
line are given by following equations:
(3.117)
(3.118)
(3.119)
(3.120)
(3.121)
The discharge pressure and speed of the lift air blower are
given by:
(3.122)
(3.123)
so
3.7. Catalyst Circulation
The catalyst circulation model was taken from Georgakis
et al. (1990). Circulation of spent and regenerated catalyst
is modeled as single phase flow governed by simple force
balances. Factors which affect catalyst circulation in a real
FCCU, such as concentration of carbon on regenerated catalyst
and injection of steam at various points in the U-bend, have
been ignored in this model. Because the dynamics of catalyst
circulation lines are orders of magnitude faster than other
dynamic elements in the FCCU system (notably the regenerator
thermal dynamics) , a pseudo-steady-state assumption is applied
to the force balances on the spent and regenerated catalyst
lines. Constant friction factors are assumed for each
circulation line.
The force balance on the regenerated catalyst U-bend is
given by
dvrgc (3.124)
dt
where
1 (3.125)
Mrgc = [ W'"'
-.-
+ ( Aubendrge Lubendr(IC Pc) ] ( -g ) '
(3.126)
51
(3.128)
The force balance on the spent catalyst circulation line is

written as
dvsc
-= (3.129)
dt
where
(3.130)
(3.130)
(3.132)
(3.133)
3.8. Numerical Procedure and

Validation of Simulator
While developing the FORTRAN program for the dynamic
simulator of integrated FCC unit, the Amoco model was used

first for the reactor-riser section of the FCC process. The
set of time dependent nonlinear ordinary differential
equations was integrated using fourth order Runge-Kutta
method. It was found that the step-size above 0.3 seconds
makes the simulator unstable. The set of spatial differential
52
equations in the regenerator was integrated using the Explicit
Euler method with a step size of 0.15 ft. To increase the
speed of computation, the regenerator spatial differential
equations were integrated after every 3 steps in time domain
without loosing any simulation accuracy. It was observed that
the computation was six times faster than the real time on a
50 MHz IBM compatible machine with a 40486 co-processor. The
simulator was validated by comparing the open-loop responses
obtained with those given in Georgakis et al. (1990). The
dynamic responses obtained using this FORTRAN simulator for a
0.05 step change in fresh feed coke formation factor are given
in Appendix D. These dynamic responses can be compared with
those given in Georgakis et al. (1990).
After this step, the Amoco model for the reactor riser
was replaced by the Lee model for the reasons explained in
section 3.3.2. The riser spatial differential equations in
the Lee model were integrated using fourth-order Runge-Kutta
method with a step size of 0.01 non-dimensional length of
riser. Then the modification such as time lag associated with
catalyst flow in reactor discussed section 3.3.2, analyzer
delay for regenerator stack gas oxygen concentration
measurement were added. This simulator was then used for the
control and optimization studies. The initial conditions for
state variables and initial valve positions are given in
Appendix B.
53
3.9. Analysis of Open-Loop Responses
3.9.1. 2% Positive Step Change
in Coke Formation Constant
The primary disturbance to the FCCU is unmeasured feed

composition changes. Since the feed composition is not
modeled in detail, a feed composition change is simulated by
changing the coke formation constant. As positive step change

is given in the rate constant of coke formation, carbon level
on the spent catalyst goes up. Therefore, more carbon is
burnt in the regenerator and regenerator temperature
increases. Increase in regenerator temperature increases the
riser temperature. Sudden increase in carbon on spent

catalyst increases coke concentration on regenerated catalyst.
Since the rate of carbon formation decreases with increase in

carbon on regenerated catalyst, carbon on spent catalyst
decreases for a moment. But as riser temperature increases,
coke formation rate increases and carbon on the spent catalyst

again starts increasing slowly. Also as regenerator
temperature increases, rate of carbon burning increases and
carbon on regenerated catalyst decreases slowly (Figures 3.11
to 3.15).
3.9.2. 2% Negative Step Change

As negative step change is given in the rate constant of
coke formation, carbon on spent catalyst decreases.
Therefore, less carbon is burnt in the regenerator which
54
decreases the regenerator temperature slowly. Decrease in
regenerator temperature decreases the riser temperature.
Sudden decrease in carbon on the spent catalyst decreases the
carbon concentration on the regenerated catalyst. Since rate

of coke formation increases with decrease in carbon on
regenerated catalyst, carbon on spent increases for a moment.
But as the riser temperature decreases, rate of coke formation

decreases and hence carbon on spent catalyst starts decreasing
slowly. Also decrease in regenerator temperature decreases

rate of carbon burning. Bence, carbon on regenerated catalyst
starts increasing. This also decreases rate of carbon
formation and assists the decrease in carbon on spent
catalyst. Also with decrease in riser temperature, oil
conversion decreases. The comparison of this open-loop
response with open-loop response for 2 % positive disturbance

discussed above clearly shows the nonlinearity of the process
(Figures 3.16 to 3.20).
3.9.3. 20°F Positive Step Change

in Fresh Feed Temperature
As positive step change is given in the fresh feed
temperature, riser and regenerator temperature increases.
Since the feed temperature directly affects riser temperature,
the riser temperature is affected sharply when a disturbance
enters the system and then the rate of increase in riser

temperature slowly decreases to reach a steady-state. Since
the coke formation rate in the riser is proportional to the
55
riser temperature, the carbon on spent catalyst increases.
Due to a sudden increase in the carbon on spent catalyst, the
carbon on regenerated catalyst increases since the rate of
carbon burning is not sufficient to cope up with the increase

in coke arrival to regenerator. But as the regenerator
temperature starts increasing, the rate of carbon burning
increases and hence the carbon on regenerated catalyst starts
decreasing until it reaches a steady-state. Due to the
increase in the carbon burning, the oxygen consumption in the
regenerator goes up and thus the flue gas oxygen level
decreases. The gas oil conversion increases due to higher
riser teJperature (Figures 3.21 to 3.25).
3.9.4. 20°F Negative Step Change

As negative step change is given in the fresh feed
temperature, both riser and regenerator temperature decreases.

The riser temperature decreases fast first and then slowly
approaches a steady-state. Since the rate of coke formation
is proportional to riser temperature, carbon on the spent
catalyst decreases. Due to decrease in carbon on spent
catalyst, less coke goes to the regenerator and hence carbon

on the regenerated catalyst decreases. But as the regenerator
temperature decreases, the rate of carbon burning decreases.
Therefore, although the coke coming to regenerator from the

reactor decreases, the net effect is increase in carbon on
regenerated catalyst. Due to decrease in rate of carbon
56
burning, the oxygen consumption in the regenerator goes down
and thus the flue gas oxygen level increases. Also gas oil
conversion decreases due to lower riser temperature (Figures
3.26 to 3.30).
3.9.5. 2 % Positive Step Change

in Feed Rate
Since the temperature of feed is much lower than that of

regenerated catalyst and cracking is an endothermic reaction,
positive step change in the feed rate decreases the riser
temperature. This causes both conversion and coke formation
in riser to go down. Due to decrease in coke formation,
carbon on spent catalyst decreases. The regenerator
temperature goes down due to lower riser temperature and less
heat released by the carbon combustion. The carbon on
regenerated catalyst first drop due to drop in carbon on spent
catalyst. But as regenerator temperature decreases, rate of
coke burning decreases and hence carbon on the regenerated
catalyst slowly increases to reach a steady-state. Since less
oxygen is required to burn off the coke, flue gas oxygen level
increases (Figures 3.31 to 3.33).
3.9.6. 2 % Negative Step Change

in Feed Rate
A negative step change in the feed rate first increases

the riser temperature. Increase in riser temperature causes
increase in conversion and increase in coke formation since
57
rate of coke formation is proportional-to riser temperature.
Therefore, carbon on spent catalyst increases. Regenerator
temperature increases due to higher riser temperature and more
heat released by coke combustion. The carbon on regenerated

catalyst first increases due to increase in carbon on spent
catalyst. But as the regenerator temperature increases, the

rate of carbon burning increases. Therefore, more and more
coke is burned off and thus carbon on regenerated catalyst
drops slowly to reach a steady-state. Increase in coke

combustion consumes more oxygen, hence the flue gas oxygen
level decreases (Figures 3.34 to 3.36).
3.9.7. 30°F Positive Ramp Change

in Ambient Air Temperature
The throughput capacity of FCCUs is affected by ambient

conditions. Higher ambient temperature decreases the capacity
of regenerator air blowers, thereby decreasing the
concentration of excess oxygen in the regenerator bed. For
simulation purpose, 30°F positive ramp is applied over 60
minutes, normally this level of disturbance would occur over

6 to 8 hours, as part of normal daily cycles of ambient
conditions. During the ramp period, oxygen concentration
drops, carbon monoxide concentration rises and regenerator
temperature increases (Figures 3.37 to 3.39).
58
Spill
air
Downstream
Separators
Gas
Oil
Figure 3.1 Schematic of Model IV FCC Unit
59
TO REACTOR
~-o:::------t----,-----=:=- RISER
T3
1-----~
vs FS
Figure 3.2 Feed Pre-Heat System
60
P4
CYCLONE
Tr
REACTOR
RISER
STEAM
FtED FROM
--====:J"E~-- PR [H [AT SYSTEM
SPENT CATALYST
TO REGEN ERA T0 R.=-E;..__~.__-=:====-../
Figure 3.3 Reactor
61
-m
0
0.013
-...,
0 0.0125
m Lee Model
:>. 0.012
.,
.-4
.,
~
0.0115
0
..., ·-~
~
Q)
~
0.011 -----~·
a)
~ 0.0105 ········---·- , ·--------

0 Amoco Model
~
0
.Q 0.01 -··--·---·
.,
~
0
0.0095 T I I I
6.3 6.55 6.8 7.05 7.3 7.55 7.8 8.05 8.3

Catalyst .to Oil Ratio
Figure 3.4 Comparison of Reactor Riser Models

(Effect of Catalyst to Oil Ratio
on Production of Coke)
62
Amoco Model
=. 777 -----··············-·····-··--··--·-··---·--··---····-·····-·-·-·--··--"x-- -----···
Q)
~ 774 ----- - ··--······ ··--·--·· ---------------· - ··----~-- -----·-·

~ Lee Model
~
Q) 771
E-t
....
Q)
m
~ 768 ················--·················-···· .. ·-------·····------------········-·······--------··------
765+-----~------~----~------~----~----~
390 400 410 420 430 440 450
Feed Oil Temp (K)
Figure 3.5 Comparison of Reactor Riser Models (Effect

of Feed Temperature on Riser Temperature)
63
790~------------------------------------~
Amoco Model
785 -----------------------------------------------------------------------------
- 780 - -- --------------------------- -
§ 775 -----··-------··---··· -·--···--·-------- . ------~ -----------

e 770 --------------------------------------------------
Cl)
Lee Model
--------------------------------------------------------------
~
Q)
E-4 765 ---------------------------- ----- -----------------------------------------------------------------------------------
~
m 76o --------
·ri
~
755 ---------------------------------------------------------------------------------------------------------------------------------------------------------------
750+-----.----.-----.----.-----.----.-----.--~
6.3 6.55 6.8 7.05 7.3 7.55 7.8 8.05 8 .3
Catalyst to Oil Ratio
Figure 3.6 Comparison of Reactor Riser Models (Effect of

Catalyst to Oil Ratio on Riser Temperature)
64
VIJ
Vl2 TO VAPOR
RECOVERY UNIT
MAIN
-----=~ FRACTIONATOR
FROM
REACTOR
Figure 3.7 Main Fractionator and Wet Gas Compressor
65
1------3~ SlACK CAS
02 CONC. (%)
Tcyc
CYCLONE
P6
Treg
REGENERATOR
STANDPIPE
REGENERATOR
LIFT PIPE ----::;;.~ REGENERATED CATALYST
TO REACTOR
AIR FROM
~=====::]........::=---SPENT CATALYST
LIFT AIR BLOWER '- FROM REACTOR
Figure 3.8 Regenerator
66
VENT
AT~OSPHERE
F6 F7 COMBUSTION AIR
TO REGENERATOR
P2
Figure 3.9 Regenerator Combustion Air Blower
67
VENT
SETPOINT
STEAM
V8
ATMOSPHERE
F8 F9 LIFT AIR
TO REGENERATOR
P3
Figure 3.10 Regenerator Lift Air Blower
68
~ 1290~--------------------------------------~
O'l
Q)
-o
- 1284 ··---·--·-·---·-····--------·-··--·-·-··------··-··--·-·······--·······-······-·--··---·
Q)
J..t
:::t
+'
~ 1278 ··--·--···-··-···-·-·--···---··· ·--·-··-··-···-·········------·-···-·---·····-··-··············-···-·····-·--·
Q)
0..
sQ)
8 1272 ·····-···-···· ····-·-···-··-··-················-···········-··-··················--·-·--···-··········--······-··-····-·-······--·--·-·-·-···
~
0
+'
~ 1266 ················-······-·····-······-····-··············-·-·-·-·-····-······--·······--·----·········-·············--················--=-------
Q)
~
Q)
O'l
~ 1260+-------~----~------~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)
Figure 3.11 Open-Loop Response of Regenerator Temperature

Constant
69
1000~-------------------------------------.
rz.t
O'l
cu 998 ······-···-------------------------·····-···-···------------------······-··-·····--------------------······--
-
"0
Q)
~
...,::l 996
cU
~
Q)
~
Q)
994 ··-···· ----~------------------------------------------------·-----------------------------------·····-·····-··········------ ---- ----·-
8
~
0
4J 992 ··········------···········--···························---····-····----························-···-···-·······-······-··········-·····-·················-··
u
«!
~
990+-----~------~----~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)
Figure 3.12 Open-Loop Response of Reactor Temperature

Constant
70
~ 0. 0085~--------------------------------------~
,...f
'.Q
,...f
- 0.0082 ··-··-·········-···-·-···-····---·-----·-··-······--··-··--·-··-··-·------····-·-·····-······
+)
m
:>,
,...f
~ 0.0079 J····-···-··········-··--·················-···--··-·······································-·-···-·············-················--··········-······
u
+)
~ 0.0076 -············-·········-··-···············--··-··························································-····························--·········-···-····-··········
~
Ul
r::
0
0.0073
r::
0
.,u
.Q
H
0.007+-------~
. -------.-,-----,,.------.,------~------~
0 50 100 150 200 250 300
Time (Minutes)
Figure 3.13 Open-Loop Response of Carbon on Spent Catalyst

for 2% Positive Step Change in Coke Formati on
Constant
71
.Q
r-i
~ 0.00095-r-------------------~
-
r-i
...,
:, 0. 00092 ·-··-····--·--···-··-·········--·-·····-···········-·········-···--·-·····-----·-···--····-·····-··-·-······-··········-·
.,
r-i
+»
· -····-~---·······-·············--··-··-···-·--······--·--·--·-·--····
«S
u 0.00089-
"0
.,
Q)
+»
~~~--~------~-
~
Q) 0.00086 ~--········-············-·······-·······················-··-··············································-····-········-·················-·-············
aQ)
0'1
~ 0.00083
a
0
a
~ 0.0008+-----~------~------~----~~-----,~----~
.,u
~ 0 50 100 150 200
Time (Minutes)
250 300
Figure 3.14 Open-Loop Response of Carbon on Regenerated

Catalyst for 2% Positive Step Change in Coke
Formation Constant
72
-c::
~ 2.-----------------------------------------------~
0
-r-f
~ 1. 7 ------------------------------------------------------------------------------------------------
«S
$.4
~
c::
Q)
u 1.4 ----- ------------------------------------------------------------------------------------------------------------------
J::
0
u
J::
Q) ·········--·--··--···-·····-········-····----········--···········- ·-·····------··
0"1
>,
X
0
~ 0.8 ---------------------------------------------------------------------------------------------------------------------------------------------------------------
(!)
Q)
::s
r;! 0.5
0 50 100 200 150 250 300
Time (Minutes)
Figure 3.15 Open-Loop Response of Flue Gas Oxygen

73
~ 1280~------------------------------------~
0'1
cv
"0
CD
~
...,:::1
~ 1260 .........!._·-········-·--········· ·····-··------······-··--·-······-·····------- --···----·······---··-···--·--·-
cv
~
acv
E-t 1250 ·······-·--····-·-·-··-····--·····---········-··-·····--··---·-··-·---· ·····-·--·-···-·---······-·--···-··----
~
...,
0
~ 1240 ·········-···············-·······-·····-···········-·-·······--·-··-·-·······-···-··--··-···--·-····---·····--·····-····-----·--·-··-··
cv
~
cv
0'1
~ 1230+-----~------~----~----~------~----~
0 50 100 150 200 250 300
Time (Minutes)

Constant
74
1000~------------------------------------~
f&4
0\
996 ·-··--···-··-·-----------·----·---···-··-·----·····-------·--··-··--···-·····-···-···---
-
Q)
"0
992
988 ················-··························----·-·········-·-··
984
980+-----~------~----~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)

Constant
75
~ 0.0082~----------------------------------~
r-t
.......
..Q
-
M
...., 0. 008 ·--··--···-·--·-··-·-·-·-··---·---··--·--·····--····-··--···-·-········-····-·-······················-··········-·····

Gl
>.
r-t
tl 0. 0078- ':::::::; ········-·-··--·············--··············-··················--·····-························-···········-·-········-······················-·
«S
tJ
....,
; 0.0076 ···········-···-·······-···-··-····-·-··-············· ··········-·····················-················-····················-··················-····-
0..
Ul
J::
0 0. 007 4- -·············---··············-··-··-··-······-·--··-····-·-····················-········-····-·-··········-··-··--···-·······--·····-··········-···
~
0
..Q
~
«S 0.0072+-----~------~----~----~------~----~
tJ 0 50 100 150 200 250 300
- - - -
Time (Minutes)
Figure 3.18 Open-Loop Response of Carbon on Spent Catalyst

Constant
76
.Q
,......
:0
,......
0.001~------------------------.
-....,
:. 0.00094 ··-···--···-·------------------------·-···-·······-·-····--····-·-···--·········--·······----·
,......
....,ItS
cu
u 0 .00088 ······-·······--······-·-·-------·········
"0
....,Q)
cu
H 0. 00082 ···············-·············-··················--···-····-··-·-···-··········-······-·-········-···················-····································
Q)
s::=
Q)
Ol
~ 0. 00076 ·················-····-··-··-····-···---······--··-·····--······--····-·-·-··-······-·-················-······-···-·······-······-····
0.0007+-----~----~------~----~----~----~
0 50 100 150 200 250 300
Time (Minutes)
Catalyst for 2% Negative Step Change in Coke
Formation Constant
77
-
~
~
3~----------------------------------------------,
0
·ri
+) 2.6 ···-···--·-·····--··-···-·-·--··--··---·----·····-··-···--··--·-------·-·-··----·····
Cd
1-4
+)
~
Q)
0 2.2 ···-·-··--·········-·········-··---····-··--···
~
0
u
~
Q)
0"1
1. 8 ·--···- ····-·-·-·········-················-·-·········-·--····--···--·········--··-··-·-·-······-···-·--·····--·······-·-···----···--····
>,
~
0
~ 1. 4 ····················-···············-······-········-··-·--····-···-··-··········--··-·--·-·--·-·---···············-··-··-·-·-··---··-·----··
C)
Q)
::s
r-1
1+-------~~-------.-------.------.-------.-------4
~
0 50 100 150 200 250 300
Time (Minutes)

78
1005·~--------------------------------------.
-
fz.t
Ol
~ 1002 ·-·-·-··-··-···-·····-·--··-··----------··-- -···---·-··-··-·--·---······-··-·-···-··--·--·
-
Q)
J.4
::s
~
Cd
J.4
Q)
~
Q)
996
8
J.4
t
0
993 ··-··--·-······-···-····---····-····--·-·····---··-··----·····-·-··········-······-·······--·--···-··-·-··-·--·-······-··-········
Cd
~
990+-----~~----~------~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)

for 20 deg F Positive Step Change in Fresh
Feed Temperature
79
1300~------------------------------------~
0\
CD
~ 1292 -------------------------------------------------------------------------------------------------
CD
$-4
:::J
+>
~ 1284 --------------------------------------------------------------------·=---....
-- - =----=----:::7:
·---=-
----=---------------------------
CD
~
Q)
8
$-4
0
+>ctJ
$-4
Q)
1268 ------------------------------------------------------------------------------------------------------------------------------------------ -------
r::
Q)
0\
~ 1260+-----~------~----~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)
Figure 3.22 Open-Loop Response of Regenerator

Temperature for 20 deg F Positive
Step Change in Fresh Feed Temperature
80
-
,..Q
~
0.0085,--------------------~
........
..Q
-
~
...., 0.0083- ·-···· ---------··------··-··-···-·------·-- ·····-·····-·······--······-··--··-·----·----··--·

co
>.
~
~ 0.0081 ---~--------····-·- ·····-- - ---·--·--·-· --·--··--· ··-·-·····-· ·· ··--···-----· -- ---------· - ······· ···- -- -- --· - ---- --
«S
u
....,
a 0.0079~······
cv
0..
(})
a
0 0.0077- -·····-···-·-··-··- ··-··-·-···--·-···-·-·············-·····-··-·······-················-····················--·-···-··-··-··-···---·······
a0
..Q
).f
«J 0.0075+-------~,----~,-------~,----~T-------~~----~
u 0 50 100 150 200 250 300
Time (Minutes)
Figure 3.23 Open-Loop Response of Carbon on Spent Catalys

for 20 deg F Positive Step Change in Fresh
Feed Temperature
81
.Q
r-i
........
.Q 0 .0 0 0 9 5 . . . . - - - - - - - - - - - - - - - - - - - - - - - - ,
-
r-i
----·- ·-··---------------·-·- ----------·········-··-···-···-····-···-----------····------
0. 00087 ······-· ....... ·--···-··-····--··-·--·-··········-··-······-············-··············-······-···········-·············-··--·---

"t..
Q)
.&J
ro
~ 0 .00083 ····-····-·········································-···········-··············-··············-·······················-····-····················-······-···
s::
Q)
Ol
Q)
p:: 0 .00079 ··················-························-·····-··-··-····-·-·-···-····----····-··--·····--·········-··········-········-·····-·····--
J::
0
8 0.00075+-----~------~----~----~------~--~
0
.Q
S-f 50 100 150 200 250 300
co Time (Minutes)
u
Catalyst for 20 deg F Positive Step Change
82
2.------------------------------------------------,
-r::
~
0
~ 1. 7 ------------------·--------------------------------------·---------------------------------------
«S
....,~
r::
Q)
u - - ---- -------------·····----------------------------------------------
r::0
u
r::
Q)
Ol
1.1
>,
~
0
m 0.8 ·----····-········-····-------------------------------·------·---------------·----------------·-··-····-····--------------·------------------·-
«S
(!)
Q)
::J
r-i 0.5+------.-------.------~----~------~----~
rz.. 0 50 100 150 200 250 300
Time (Minutes)

Concentration for 20 deg F Positive
83
1000~------------------------------------~
IZ-4
Ol
995 ----·---··-··---·····-··-······-··-·········---··-·······-·----···-·······-·····-·········---··------·-··
-
G)
"'0
G)
~
~ 990
~
a:s
~
G)
m 985 ··--·---------·--··-····-····-···-···--·-·--·····-········-···-·············-···· ---=

····-=····=-=
···-::.:.::.
·····::.::.:
····=····-=····=····=·
·········-------···-·····--
8
~
0
t; 980 ···----····----·········-·--·--·--·-·--··········-··-··········--·---····--··--·-·-··········-·-·----····-···--------
a:s
~
975+-----~------~------~----~------~----~
0 50 100 150 200 250 300
Time (Minutes)
Figure 3.26 Open-Loop Response of Reactor Temperature for

20 deg F Negative Step Change in Fresh Feed
Temperature
84
~ 1280~------------------------------------~
t}\
Q)
ro
- 1270 ···--···· ··-······-··-·····-··--··-·-·-···-·-··············--····-····-··-·····-····-·····-········-······--···---··-···---··--·
Q)
~
::s
.f.)
nJ
~
1260 ·--·····-···-·-····-···-···-· ··--·--······-···-·······-··-·-···-··········-··········-··········--··--·····-·······--·--·····--···
Q)
~
Q)
E-t 1250 ················-·····-·--·······-··-·····················-·····--····················-··········
~
0
.f.)
ctJ 1240 ·····················-··-·-··-······-········-················-·············-······--·····-····················-··--························-··--·······-···
~
Q)
~
Q)
t}\
~ 1230+-----~------~----~------~----~----~
0 50 100 150 200 250 300
~---------------------
Time (Minutes)

for 20 deg F Negative Step Change in Fresh
Feed Temperature
85
.Q 0 .0 0 8 2 . - - - - - - - - - - - - - - - - - - - - - - - .
r-t
.......
.0
-
M
0. 008- ··-··-·····-----··········-··-···········-·····---···-···-··--···-··-----··-····-·--··---··- - - - - ·- ·
0. 00 7 6- ·-·································-································-··········-·········-·····--·-······-····-·--······--········--·-············--···
~
0 0. 007 4- ····--·--············-·-······-··············-·-····-·····---·····-·········-····-··························---·-·-···········-···-·····-··-
~
0
.Q
~
~ 0.0072+-------.-----~-------,-,----~,-------Tr-----~
u 0 50 100 150 200 250 300
Time (Minutes)
Figure 3.28 Open-Loop Response of Carbon on Spent

Catalyst for 20 deg F Negative Step
Change in Fresh Feed Temperature
86
.Q
r-t
:0 0.00105.-----------------------.
-
r-t
-1-l
m
>. 0.001 ··--·--··--··-----------·-····--·--······ -·--··----·--- - - - - - - - - - -
r-t
«S
-1-l
«S
u 0 .00095 -························-···-·-····----······-·-·--····-··----···-···········-····-·----------------
"'0
Q)
-1-l
«S
~ 0 .0009 ············---··-····-···-··-··········· ··-········-·-··-···-···························-··-········-·········--·········-·············-···
J::
Q)
O"l
~ 0 .00085 ··--··-·-··---------····-··············--·--······-··············-···-···············--···--·-·······-···-----
r::
0
J::
~ 0 . 0008+-----~------~----~------~----~----~
~ 0 50 100 150 200 250 300
«S
u Time (Minutes)
Catalyst for 20 deg F Negative Step Ch ange
87
-a
~ 3~--------------------------------------------~
0
•j«S 2.6 ···-·--·-- - ---- -······- -------------·-··- ·····---·--·---·-··-········-··-··-·
J..4
~
~
Q)
0 2. 2 ··-······-···-··--·-·- -·-···--·-···--··-····--····-··-····--·- ······
~
0
u
~
Q) 1.8 ·····--···-············
Ol
:>,
~
0
~ 1 .4 ·-·········-·-·····- ···-···-····-··--··-········---····-·--···--·-·······················-····-·-········----···········-····-·-···········-····-····
{!)
Q)
::s
...... 1+-------.-------~-------.------~-------.------~
~
0 50 100 150 200 250 300
Time (Minutes)

Concentration for 20 deg F Negative
88
1000~------------------------------------~
fz-4
0'1
Q) 996 ··-·-·-···-··-- ·····---···-·-·-·--·-··-·-···--··········-·-····················-·- ·····-·······- ·-·······-····-·········-·
-
"0
Q)
~
:;j 992 ·--·-···· ······-····-··-··--···-·-·········-······-························-···········-·····-····································-·····················-··
"'-'cu
5-f
i
Q)
988 -····--··········-··-·-·--·-···---···-··········
8
5-f
t0
984 ·--····-··---········-·-···-···--···-··-·-···-······-··---·-······················-·······················--····-·····--·····-···-········
ctS
~
980+-----~------~------~----~------~-----4
0 50 100 150 200 250 300
Time (Minutes)

for 2% Positive Step Change in Feed Rate
89
~
1280~----------------------------------~
0'1
Q)
"0
- 1270 ·········· ·--···········------·---··-··--·--·-··--···--······--·-······-···-·-·····--·--···-·----····-···---·---·-·------·--···-··
Q)
J.f
::s
+J
cu 1260 ·--···--······---·--··-·--·-·-······
J.f
Q)
p.
J:i
Q)
8
J.f
0
+J
cu
J.f
Q)
1240 ···---·-·----······-----·---··-·-··---·······----··-·-·······-··-··--·-···················-··················---···-···--·--·-----·--·
Q
Q)
Ol
Q)
p:::
1230+-----~------~------~----~------~----~
0 50 100 150 200 250 300
Time (Minutes)

for 2% Positive Step Change in Feed Rate
90
-
~
~
3~--------------------------------------------~
0
·r-i
~ 2. 6 ·····-·-···-······················-·····-·-···--··----·-······-·····-·····································-----···--··-······-·-···-·-··-------··
$.1
.1-J
~
Q)
g 2. 2 ···········-·-···-····--····--··-····-·-··-·-····-·-······-·················
0
u
~
~ 1.8
>t
><
0
~ 1.4 -··········-···-·······-······--····--·--····················-·--··········-··················-····················--·········-···-····-·····-··············--·
CJ
Q)
::I
~ 1+--------.------~--------~------~-------.------~
fz.t 200 250 300
0 50 100 150
Time (Minutes)

in Feed Rate
91
1005~------------------------------------~
-
rz..
Ol
CD 1002 ·············--·-··························-···········--·-··················--·······-··········--···---·····-······························-·······-·-····
-
"tj
999 ·---·····-· ··--······-·········-··---·--·---····-···········-··=···-·::.::,::··

-··=·· ···~
··· - - =···=···
·-=···-·=·····==·
= ···-==··
··-················-··········
996
993
990+-----~~----~------~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)

for 2% Negative Step Change in Feed Rate
92
-rz.. 1290~----------------------------------~
Ol
cv
"U
- 1284 -·-·--·····-······--····--·--···-·-·-·--·-··------·-·-···---·-·-··-·---····-············--··-·-·-··-··-····--·
cv
J..r
....,::s
«S
....
1278 ·--·-·-·-·- ······-···-·--·-
cv
~
cv
E-t 1272 ····-·--···· ··-···············-·····-·························-·······-······-·············-··-·-·-·······················-············-········-·-····-·······
J..r
0
....,
«S
J..r 1266 ·············-·-·-·······················-·-·····-·-·········-·····-······-················-·········--·····················-···················-···········-····
cv
J:::
Q)
Ol
~ 1260+-----~------~----~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)

for 2% Negative Step Change in Feed Rate
93
-
-
. .
1. 7 ······-···-··------·····-·-··-·-···---·-··--···--···-···---·----········-·--··-··------·····-······--·-·---
•••• • •••·•••--•• • - • • • · - - - - · - - • • • • - - • • • • • ••·- • •••••u • - - • •• • • • - - • - - •• • • - • • • • - - • • - • • - • - • • • - - - - • · • - - • • • - • • -· • • • • • ••••-•••-•
-··-····---------------·---------····-······························-----------····
-···=----::::
-.. -=--.:::-:
....::::
....:::: ....=
....:::: ....=...,..,.- - - -······················-·····
~ 0. 8 ···-·······--·······-······--···············--······-··-···········-·-·················-··············-···························-··············-·····················
C)
(J)
:;j
~ 0.5+-------~------~------~------~------~------~
0 50 100 150 200 250 300
Time (Minutes)

in Feed Rate
94
-J:::
2~------------------------------------~
0
·ri
+l
«S
1. 7 ···-····························-·······--····--·-········-···-·····-·····-··-····-··-··-·--·-·--···············--··········-·---···-·-·
~
.&J
J:::
cv
0 1 .4 .... ······-···-····················-·······--····-··-···········--·······-····-···········---········································-················-·····-···-·····
J:::
0
u
J:::
cv 1.1 ·····-··················-·····················-·········
0'\
:>,
><
0
til 0.8 ··············································-····································- ····--·······································-··········-······-··············-·-····-·-
«S
~
cv
:::s
~ 0 . 5+-------~------~------~------~------~------~
rz.. 0 50 100 150 200 250 300
Time (Minutes)
Figure 3.37 Open-Loop Response of Flue Gas Oxy gen

Concentration for 30 deg F Positive
Ramp Change in Ambient Air Temperature
95
~ 1300~------------------------------------~
Ol
Q)
"0
- 1292 ··---····-······-··········--·-····--·--·-·-·····-·····---·-··-----·········-······-·-·······--------············-···--·--·--···-··
Q)
~
::s
....,
~ 1284 ·--·····-·-····-········-·····-··-········--······-··-···-···--···-··-····-······-··-····-······-················-······-············----···-·····
Q)
~
Q)
E-t 1276 ·················-····-· ······-·-············-················-················-·····························-·········-········-··-·····-··-···-···-····--·
~
....,0
ctS
~
Q)
1268 ····-·-····-·····--···-··············-··············-·-··-----·-·-·······-·············-···-···································-·······-··········-····
~
Q)
Ol
~ 1260+------,.-----~------~------~----~------~
0 50 100 150 200 250 300
Time (Minutes )
Figure 3.38 Open-Loop Response of Regenerat or

Temperature for 30 deg F Positi ve
Ramp Change in Ambient Air Temperature
96
1000~--------------------------------~
-
0"1
Q) 998 ----------------------------------------------------------------·--·----·-·········---···-····-·······-·-·-····--·-·-·----··
-
1j
996
994
992
990+-----~------~----~------~----~----~
0 50 100 150 200 250 300
Time (Minutes)

for 30 deg F Positive Ramp Change in
Ambient Air Temperature
97
CHAPTER 4
NONLINEAR PROCESS MODEL BASED CONTROL
Control techniques that directly use a nonlinear model to

make control decisions are referred to here as nonlinear
process model based control (nonlinear PMBC) methods.
Recently a number of nonlinear PMBC methods have been
developed that provide nonlinear feedback and feedforward
control using a nonlinear model. One of the first nonlinear
PMBC methods developed was nonlinear internal model control
(nonlinear IMC, Economou et al., 1986). Nonlinear IMC relies
upon an iterative integration of the process model to
calculate control action in conjunction with the IMC structure
and a nonlinear filter. It uses a setpoint bias to remove
offset due to process/model mismatch. Generic model control
(GMC, Lee and Sullivan, 1988) uses an explicit formulation for
its control law along with an integral term to remove offset.
Also, the Incremental Model Parameterization On-Line (IMPOL)
method (Riggs and Rhinehart, 1991) can be used along with GMC
control law to remove the process/model mismatch. Riggs and
Rhinehart (1988) compared nonlinear IMC and GMC for a wide
range of CSTR and heat exchanger control problems and found
essentially equivalent performance. Parrish and Brosilow
( 1988) developed another nonlinear PMBC method that they
called nonlinear predictive model control (NLPMC) in which
process/model mismatch is accommodated by adjusting unmeasured
98
disturbances. The advantages of nonlinear PMBC methods come
from the nonlinear process description and the nonlinear
decoupling provided by the process model. This work involves
use of nonlinear process models with Generic model control as

a control law. Nonlinear PMBC has been successfully applied
to processes such as coal gasification (Pandit et al., 1988),
pH (Choi et al., 1988; Williams et al., 1990; Mahuli et al.,

1992), commercial distillation (Riggs et al., 1989),
wastewater neutralization (Rhinehart et al., 1989) and

supercritical fluid extraction (Ramachandran et al., 1992).
4.1. Approximate Controller Models

The PMBC strategy uses nonlinear approximate models
directly to calculate the control action. These approximate

models are not rigorous simulators but should contain the
major phenomenological characteristics of the process that
they represent. Nonlinear approximate models provide a

nonlinear picture of the process and have a considerably wider
range of effective process description than linear process
models. Riggs and Rhinehart ( 1989) present a review of
approximate nonlinear models for various chemical processes.
Another criterion while selecting an approximate process model

is its computational efficiency. The approximate model must
be computationally efficient. To be computationally rapid,
the approximate model should be considerably simpler than the

real process; but, to be .functionally useful, it should
99
mechanistically account for the dominant phenomena of the
process. Further, numerical implementation of the approximate
model must be reliable in spite of bias and random errors in
measured data. Also, the model should be simple enough to be
understood by plant engineers. Therefore, a successful
approximate model is a judicial compromise between simplicity
and function.
4.1.1. Approximate Model for Reactor

In the reactor, nonlinear process model based controller
is used for the control of reactor riser temperature.
Therefore, the approximate phenomenological model is developed
for the reactor riser temperature. A macroscopic steady-state
energy balance is used as an approximate model for this
purpose. The approximate model is given by following
equation:
(4.1)
where
= catalyst circulation rate, lb/s,
= specific heat of catalyst, 0.24 BTU/lb/°F,
Foil = gas oil flow rate, lb/s,

= specific heat of gas oil, 0.75 BTU/lb/°F,
= reactor t emperat ure, oF,
Treg
- regenerator Tempera t ure, oF,
= temperature of feed to reactor riser, °F,
Tfeed
100
Conv = conversion of gas oil,
AHcrack• endothermic heat of cracking, BTU/lb.
4.1.1.1. Calculation of Gas Oil Conversion

The conversion of gas oil is estimated from measured
variables using following empirical equation:
(4.2)
This equation is obtained by integrating the differential

material balance on gas oil across the length of reactor
riser. The temperature is assumed to be constant along the
length of the riser in order to obtain a closed form solution.
Also, since the carbon concentration on the regenerated
catalyst is not measured, it is lumped with constant k 1 •
These assumptions create uncertainty in the above equation,
which will lead to incorrect prediction of gas oil conversion
and hence the reactor riser temperature. To overcome this
process/model mismatch,,constant k 1 is used as an adjustable
parameter for the on-line parameterization of this approximate
model. In above equation,
COR = catalyst to oil ratio,
tc - catalyst residence time in riser, Seconds,

Tr = riser outlet temperature, °F,
1
k1 = empirical constant, s- ,
101
-1
= empirical constant, s '
Ef • activation energy of gas oil cracking reaction.
The activation energy Ef and the constants k 1 and k 2 are
evaluated by nonlinear regression using steady-state data
given in Appendix C. The constant k 1 in the above equation is
used as an adjustable parameter to remove the mismatch between
the process value of reactor riser temperature and the value
predicted by the above approximate model. This was the
obvious selection since this is a parameter with maximum
uncertainty in it. The values of k 2 and Ef are kept constant
throughout the control studies. The values of these
parameters obtained by nonlinear regression of steady-state
data are as follows:
7
0.112334 X 10 s -1 '
0.1067935 s-\
= 15533.27 °F.
4.1.2. Approximate Models for

the Regenerator
In the regenerator, the regenerator temperature and the

flue gas oxygen concentration are controlled by nonlinear
process model based controllers. Therefore, approximate

models are developed for the regenerator temperature and the
flue gas oxygen concentration. The approximate models for
regenerator are described in the following sections.
102
4.1.2.1. Approximate Model for Flue
Gas Oxygen Concentration
Approximate model for the flue gas oxygen concentration

is a dynamic macroscopic oxygen balance for the regenerator.
It is taken from Lee et al. (1985). The model assumes that:
1. The fluidized catalyst phase in the regenerator is
assumed to be perfectly mixed,
2. The kinetic equation employed for the coke combustion in
the fluidized bed regenerator is that presented by Weisz
and Goodwin (1963). Weisz's kinetic model gives
(4.3)
where
r = rate of coke combustion,
K = rate constant,
cc = coke concentration on catalyst,
y = flue gas oxygen concentration.
3. The coke composition can be represented by CBn, where n

is assumed to be 2.
The oxygen balance equation is given by,
(4.4)
where
w. = air holdup in the regenerator, lbmols,
Y. = concentration of oxygen in regenerator stack gas,

- concentration of oxygen in air entering
regenerator, 0.21,
103
RA = air flow rate to the regenerator, lbmol/s,
a • C0 2 /CO ratio in regenerator flue gas,
Me = coke molecular weight, 14 lb/lbmol,
k = rate constant for coke burning, s -1 ,

• k 0 exp(-18989/Tr•g), Tre9 in K
k0 = pre-exponential in the rate constant for coke

burning,
Treg = regenerator temperature, K,
CRc = carbon concentration on regenerated catalyst, wt%,
wreg - catalyst holdup in the regenerator, lb.

The dynamic oxygen balance equation is simplified further by
making following assumptions:
1. Air holdup in the regenerator is assumed to be constant,
2. It is assumed that concentration of C02 in regenerator
flue gas is much higher than concentration of CO,
Therefore,
C02
0 = > 1. (4.5)
co
l+l.So
(1 +o)
=1. 5. (4.6)
3. Since carbon concentration on regenerator catalyst (C~)
is not measured, it is lumped with the pre-exponential of

rate constant, k 0 •
so the approximate model for flue gas oxygen concentration
becomes,
104
W _dy
_e = ,R \(y -y ) - 1.5k0 Ye W.reg exp ( - 18989 )
a dt A i e M T (4.7)
c reg
Estimation of Catalyst Holdup in Regenerator:

The combustion air flow rate to the regenerator is given by,
(4.8)
where
F7 = combustion air flow rate, lb/s,
kc~ = combustion air blower discharge pipe flow

resistance factor, 62.94 lb/s psia0 " 5 ,
P2 = combustion air blower discharge pressure, psia,
Prgb = pressure at bottom of regenerator, psi.
Therefore, the pressure at bottom of the regenerator is given
by,
- - F,
Prgb - P;a ( k )2 . (4.9)
comb
Pressure at the bottom of regenerator can also be calculated

from following equation:
(4.10)
where
prgb
= pressure at the bottom of regenerator, psi,
= regenerator pressure, psia,
wr•g
= catalyst holdup in regenerator, lb,
• area o f regenera t or, ft 2.
105
Therefore, catalyst holdup in the regenerator will be given
by,
(4.11)
(4.12)
Since P2 , P 6 and F7 are measured variables, catalyst holdup in

the regenerator (Wr•g) can be calculated.
Estimation of Air Holdup in the Regenerator:
Density of gas in the regenerator is given by,
(520) p6
Pg = (379) {14.7) (Treg+459.6)
(4.13)
where
P6 • regenerator pressure, psia,
Treg = regenerator temperature, °F,
p9 =density of gas in regenerator, lbmol/ft 3 •
Superficial velocity of catalyst in the regenerator is given
by,
(4.14)
where
v. • superficial velocity of catalyst in
regenerator,ft/s,
F. 9 = stack gas flow rate, lbmol/s,
Fair = air flow rate to regenerator, lbmol Is,
106
Areg • cross sec ti ona 1 area o f regenera t or, ft 2.
Stack gas flow rate used in calculation of superficial
velocity of catalyst in the regenerator is calculated from

following equation:
(4.15)
where
= regenerator stack gas valve flow rating,
0.01 lbmol/s psia

0.5
,
= regenerator pressure, psia,
P.ta = atmospheric pressure, 14.7 psia.
Once superficial velocity of catalyst in the regenerator (v.)

is calculated, other dependent variables can be calculated.
Apparent void fraction in regenerator dense phase bed (ef) and
density of catalyst in the dilute phase ( Pc,dilute) can be

calculated using following empirical equations:
ef = o. 332 + o. 06 V 8 • (4.16)
Pc,dilute = -0. 87 8 + 0. 582 V 8 • (4.17)
Further, density of catalyst in the dense phase bed ( Pc,den••)

can be calculated in terms of apparent void fraction in
regenerator dense phase bed as follows:
(4.18)
where
p~
3
• settled density of catalyst, 68 lb/ft •
Then, regenerator bed height can be calculated using following
107
empirical equations.
Zbed = min[zcyc' (2.85+0.8v + Wreg-Pc,diluteAregZcyc) ( 1 )'

8
AregPc,dQllBB l - Pc,d1luts
Pc,denss
(4.19)
where
z~ = regenerator dense bed height, ft,
z~a = height of cyclone inlet, 45 ft,
Pc,dilute • density of catalyst in the dilute phase,

3
lb/ft '
Pa,denee = density of catalyst in the dense phase,

lb/ft 3 •
Once regenerator bed height (z~) is calculated, the effective
void fraction in the regenerator dense phase bed (e.) can be
calculated using following empirical equation.
(4.20)
Regenerator volume occupied by gas can be obtained as follows:
(4.21)
Knowing regenerator volume occupied by gas, air holdup in the

regenerator can be calculated as follows:
(4.22)
108
Regenerator Temperature
Approximate model for regenerator temperature is a

dynamic macroscopic energy balance for the regenerator. It is
taken from Lee et al. (1985). The model assumes that:

1. The fluidized catalyst phase in the regenerator is
assumed to be perfectly mixed,

2. The energy balance assumes adiabatic operation of the
regenerator.
The dynamic macroscopic energy balance is given by,
(4.23)
where
wreg = catalyst holdup in regenerator, lb,
CP• = specific heat of catalyst, 0.24 Btu/lb°F,
M
1
= effective heat capacity of regenerator mass,
200,000 Btu/°F,
F. = catalyst circulation rate, lb/s,

Tr = reactor temperature, °F,
Trec;r = regenerator temperature, °F,
F air = air flow rate to regenerator, lbmol/s,
Cpair = specific heat of air, 7.08 Btu/lbmol°F,
Tair -= temperature of air entering regenerator, °F,

ABco = heat of formation of carbon monoxide, Btu/lbmol,
109
ABco2 = heat of transformation of CO to C02 , Btu/lbmol,
k - rate constant for coke burning reaction, s -1
'
- k 0 exp ( -18989/Treg) 1 Treg in K
ko - pre-exponential in rate constant,
Y. = concentration of oxygen in regenerator stack

gas,
C~ - coke concentration on regenerated catalyst, wt%.
The approximate model is simplified further by making
following assumptions:
1. It is assumed that concentration of C0 2 in regenerator
stack gas is much higher than concentration of CO,
Therefore,
C02
0 = > 1. (4.24)
co
0
= 1. (4.25)
l+o
(4.26)
where
ABcok=
e
heat of combustion of coke, Btu/lbmol.
2. Since, carbon concentration on regenerated catalyst is
not measured, it is lumped with the pre-exponential of
rate constant, k 0 •
Thus, the approximate model becomes,
110
-llH koYeWr!f1exp(-18989) (4.27)
coke M T •
c reg
Estimation of catalyst holdup is explained above. The

catalyst circulation rate coming into the regenerator and
leaving it is assumed to be same. Since, in actual practice,
although the rates are different, the difference is not large.
Also, since this is an approximate model, this is a reasonable
assumption. The circulation rate of regenerated catalyst is
used for this purpose. The model equations which relate the
catalyst circulation rate to measured variables like reactor
pressure, regenerator pressure and level of catalyst in
standpipe are explained in previous chapter.
4.2. Parameterization of an
Approximate Model
Approximate models are not highly detailed models that is

they do not exactly describe the process. As a result, some
of the model parameters values must be adjusted to keep them
"true" to the process. In fact, even after initial
parameterization, process/model mismatch will usually increase
as the operating conditions are moved away from the conditions
at which the model was parameterized. In addition, certain
characteristics of the process may be time varying (e.g. ,
111
catalyst activity, heat transfer surface fouling). Therefore,
the approximate models must be parameterized regularly to be
effective for a control application. The parameters with the
highest degree of uncertainty are usually chosen as
adjustable parameters for the approximate model. The on-line
incremental model parameterization method ( IMPOL) discussed by
Riggs and Rhinehart (1991) is used for the purpose.
4.2.1. Parameterization of Reactor Model

Since this is a steady-state model, steady-state on-line
incremental parameterization method is used. The empirical
constant, k1 , in the gas oil conversion is used as an
adjustable parameter, since it has maximum uncertainty. The
difference between the reactor temperature calculated from the
model, that is, the steady-state energy balance and the
measured temperature from the simulator, is to be driven to
zero.
e = T Ipr~•• - T I.otJel • (4.28)
The adjustable parameter (k 1 ) is evaluated according to

following equation:
(4.29)
where, a is the relaxation factor to compensate for Newton's

linearization, to damp out the noise and to damp out the
dynamic mismatch of measured variables. The value of a is
112
tuned to get the best result, i.e., to get the best match
between the process and the model value of reactor temperature

dynamically and it was found to be 0.95. The partial
derivative of the reactor temperature with respect to the
adjustable parameter k 1 is calculated by finite difference
approximation using the approximate model for the reactor.
4.2.2. Parameterization of
Regenerator Models
To parameterize the regenerator approximate models,

dynamic IMPOL method is used. The pre-exponential in the
carbon burning reaction (k0 ) , and the heat of coke burning
reaction ( ABcoke) , are selected as adjustable parameters. The

heat of coke burning reaction depends upon the composition of
coke, which changes with changes in type of feed. Therefore,

it is selected as an adjustable parameter. It will also
account for the carbon monoxide leaving regenerator, which is
assumed to be zero in the development of approximate models.
(4.30)
E 2 = Trl!lg~:~Zoce•• -Trl!lg80dlll • (4.31)
Then the adjustable parameters are changed according to
following equations so as to drive the errors to zero.
(4.32)
113
(4.33)
(4.34)
= (4.35)
It can be noted from the above set of equations that the

adjustable parameter k 0 appears only in the first approximate
model. This fact leads to the decoupling of these equations
as seen in the above equations. Thus the parameter k 0 can be
parameterized using the flue gas oxygen concentration value.
Once k 0 is parameterized, ABcoke can be parameterized using the
regenerator temperature. Where, a 1 and a 2 are the relaxation
factors to account for the Newton's linearization. The values
of a 1 and a 2 are found to be 0.85 and 0.1, respectively.
4.3. Nonlinear Process Model

Based Controller
4.3.1. Generic Model Control CGMC) Law

Generic Model Control uses a nonlinear model of the
process which includes phenomenological aspects of the
process. To understand GMC, let us consider a single input,
single output (SISO) process described by the following model:
where y is the output variable, u is the manipulated variable,
114
~l = f (y' u ' .d' k.) ' (4.36)
d is a vector of measured disturbance, and k is a vector of

adjustable parameters. Assuming that y has a value of y 0 , and
it is desired to move the process from Yo to Y.P in some
interval~, then equation (4.36) can be approximated using the
forward difference approximation of the first derivative:
Ysp - Yo _ f( A Z...) (4.37)

'f
- Yo ' u' ~' A •
Then equation (4.37) can be solved directly to determine u,

the control action, if we know ~ and k. The time interval ~
is a tuning parameter. If ~ is small, rapid response will be

obtained and if ~ is large, sluggish response will be
obtained.
Since the controller model used in equation ( 4. 36) is not
exact, use of this control law will result in steady-state
offset. To eliminate this offset, Lee and Sullivan added an
integral term, resulting in the GMC control law given by
equation (4.38).
(4.38)
K1 is equal to 1/~ in equation ( 4. 37). Thus this equation can
be solved to calculate the control action.
115
4.3.1.1. Steady-State Version
of GMC Control Law
If the approximate model is a steady-state model, it
is converted into a dynamic model by assuming a first-order
dynamics.
dy
dt
= ...!.
t
(y
BB
-y)
'
(4.39)
p
where Y•• is a solution of the steady-state approximate model

based upon the current values of measured disturbances and the
current values of the manipulated variables. Combining this
equation with GMC control law yields
(4.40)
This equation can also be viewed as a PI controller being used

to select the target value of control variable. This equation
can be directly evaluated to determine Y... Then the control
action can be calculated by substituting it in the steady-
state approximate model.
4.3.2. Flue Gas Oxygen

Concentration Control
The dynamic macroscopic oxygen balance for the
regenerator is used as an approximate model for the flue gas
oxygen concentration. If the approximate model for the flue
gas oxygen concentration is represented by following equation:
'dte : f 1 (yS 1 U 1 dI PI q) I (4.41)
116
then the nonlinear control law for the flue gas oxygen
concentration is developed as described in section 4.3.1.
(4.42)
The parameters K11 and K12 are tuning constants. This equation
is solved with equations (4.44) and (4.45) developed in the
next sections simultaneously to calculate the control action.
4.3.3. Regenerator Temperature Control

The dynamic energy balance is used as an approximate
model for the regenerator temperature. It is described in
detail in section 4 • 2 • 2 • 2 • If this approximate model is
represented by following equation:
dTreg
dt =
f 2
( Trey 1 U 1
d1P1 q) 1
(4.43)
then the nonlinear control law is developed for regenerator

temperature as described in section 4.3.1. It is given by,
(4.44)
The parameters K21 and K22 are tuning constants. This equation
is solved with equations (4.42) and (4.45) simultaneously to
calculate the desired control action.
117
4.3.4. Reactor Temperature Control
The approximate model for the reactor temperature is a
steady- state energy balance equation, therefore, steady-state
version of GMC law discussed in section 4.3.1.1 is used to
develop the control law.
(4.45)
The steady-state target is substituted in approximate model

and is solved simultaneously with equations (4.42) and (4.44)
to calculate the control action. The parameters K31 and K32
are tuning constants.
4. 4. 'Basic Regulatory Controllers

These are Proportional-Integral (PI) controllers. The
control objective of nonlinear PMBC controllers is to maintain
the reactor temperature, the regenerator temperature and the
flue gas oxygen concentration at their desired setpoints.
These PMBC controllers calculate the feed temperature entering
reactor riser, the catalyst circulation rate and the
regenerator air flow rate to achieve the control objective.
These values are fed to the basic regulatory PI controllers as
their setpoints and final control action is taken by them.
Following PI controllers are used in this control
configuration.
118
4.4.1. Riser Feed Temperature Controller
The feed to the reactor riser is preheated in a feed pre-
heater before entering reactor riser. The temperature of the
feed stream leaving the pre-heater and entering the riser is
controlled by manipulating the flow of fuel to the pre-heater.
This is a Proportional-Integral controller which uses velocity
algorithm.
4.4.2. Reactor Pressure Controller

The reactor pressure is controlled by manipulating the
wet gas compressor suction valve (V11 ) , which changes the flow
through the wet gas compressor suction valve. The setpoint to
the reactor pressure controller is maintained at 33.0 psia.
4.4.3. Catalyst Circulation

Rate Controller
In Model IV fluidized catalytic cracker, the catalyst
circulation rate is altered by manipulating
reactor/regenerator differential pressure. As mentioned
above, the reactor pressure is maintained at its setpoint by
reactor pressure controller. Therefore, to alter the catalyst
circulation rate, the regenerator pressure needs to be
changed. The regenerator pressure is controlled by
manipulating the regenerator stack gas valve position (V14 ) .
Therefore, the catalyst circulation rate controller uses the
regenerator stack gas valve position as a manipulated
variable.
119
4.4.4. Regenerator Air Flow
Rate Controller
Air is supplied to the regenerator by two centrifugal

compressors. The bulk of regenerator air is supplied by the
combustion air blower, which supplies air directly to the
bottom of the regenerator. The lift air blower injects air
into the bottom of the lift pipe to assist the circulation of
the spent catalyst from the reactor to the regenerator. In
addition to this, a spill air line is available through which
the excess air from the lift air blower can be diverted
directly to the bottom of the regenerator, if desired. A
valve (V9 ) is available on this line to control the spill air
flow rate. At the base case, this valve is closed completely.
The Proportional-Integral controller maintains total air flow
rate to the regenerator by manipulating the governor setpoint
(V11ft) of the lift air blower. A separate PI controller was
also developed for the spill air line. If the desired air
flow rate is below 80 lb/s, the spill air valve is kept closed
and the desired air flow rate is maintained by manipulating
only the lift air valve. If the desired air flow rate is
above 80 lb/s, the additional air is supplied by manipulating
the spill air valve (V9 ) . This configuration was decided
after studying the behavior of these valves. The suction
valve on the combustion air blower (V6 ) is fixed fully open.
120
CHAPTER 5
CONSTRAINT CONTROL
Many chemical processes have constraints on input, state

and output variables. Therefore, the ability to handle
constraints is essential for any control algorithm to be
implemented on real processes. Although model based process
control has drawn considerable attention in process control
because of its performance, none of the techniques were
originally designed with explicit constraint handling methods.
The strategies for constraint handling within model based
algorithms has become one of the popular research topic.
5.1. FCCU Constraints

The constraints for the FCCU system have been classified
as hard equipment constraints and operational constraints.
Hard constraints represent physical limitations on equipment,
e.g. , valves fully opened or closed. Vessel maximum pressures
have also been given as hard constraints because relief valves
open above these pressures. Operational constraints represent
desired operating limits arising from economic, safety or
environmental considerations.
5.1.1. Hard Equipment Constraints

Following are the hard equipment constraints present in
the FCCU.
121
1. An upper constraint on feed flow rate,
0 ~ F011 ~ 154 lb/s
2. An upper constraint on firebox fuel flow rate,
0 ~ F 5 ~ 40 SCF/s
3. An upper and lower limits on all valves,
o ~ vi ~ 1
4. The upper limits on vessel pressures.

Regenerator maximum pressure, P6 < 39.7 psig
Main fractionator maximum pressure, P5 < 39.7 psig.
5.1.2. Operating Constraints

Following are the operating constraints present in the
Fluidized catalytic cracking process:
1. Compressor surge limits,
lift air blower, F sucn,lift > F surge,lift
combustion air blower,
wet gas compressor, F sucn,wg > F surge,wg
2. Regenerator total combustion operation.
The regenerator is operated in total combustion mode,
meaning that virtually all carbon monoxide produced in the bed
is converted to carbon dioxide before the gas stream enters
the cyclones at the top of the regenerator. Under these
conditions, carbon monoxide concentration in the stack gas is
in the ppm range. In typical operation, limits on oxygen
concentration in the bed (as measured by oxygen concentration
in the stack gas) and the bed temperature are specified as
122
operating constraints to ensure complete combustion of carbon
monoxide. If these conditions are maintained, environmental
constraints on carbon monoxide concentration in the stack gas
will not be violated. However, the operating constraints on
oxygen concentration and the bed temperature are often placed
a conservative distance from the true constraints on total
combustion, because the responses are more linear and are more
easily controlled. For this reason, two alternate sets of
constraints have been defined for total combustion operation.
The first set represents the conservative limits on excess
oxygen concentration and bed temperature, typically used to
control operation of the regenerator, while the second set
defines the true constraints on total combustion operation.
Total Combustion Constraints, Set A:
Co2,•g 2:: 1. 5 %
Treg 2:: 1265°F

Total Combustion Constraints, Set B:
Cco,. 9 ~ 350 ppm

3. Reactor/Regenerator total pressure differential,
-5.0 ~ AP0 ~ 2.0 psi.
In Model IV FCC unit, reactor/regenerator pressure

differential is used to manipulate the catalyst circulation
rate. Therefore, this constraint puts the limits on the

catalyst circulation rate that can be obtained.
123
4. Riser Temperature,
An upper constraint on riser temperature due to metallurgical
temperature limit.
Tr ~ 995°F.
5. Furnace Constraints.
An upper constraint on firebox temperature.
T 3 ~ 1700°F.
5.2. Constraint Control

In this work, the constraints on the manipulated
variables, i.e., air flow rate constraint, firebox constraints
and catalyst circulation rate constraint are considered. In
a real FCCU, various other equipment and operating constraints
are present, but the aim of this work is to demonstrate the
constraint handling procedure using a nonlinear process
models, which can be extended to other constraints.
When a constraint on a manipulated variable is hit, a
degree of freedom is lost. This means that as long as the
feed rate to the process remains constant and any disturbance
does not enter the system, control of one output variable is
lost. This further means that when a constraint on a
manipulated variable is hit, the feed rate to the system needs
to be adjusted appropriately and the desired control objective
can be achieved. This feed rate calculation can be done by
using a phenomenological process model which relates the
process variables to the feed rate. The primary objective of
124
this control strategy is to maximize the feed rate while
maintaining the riser temperature at its optimum value to get
the best product yields and while satisfying all equipment and
operating constraints.
5.2.1. Firebox Constraints

The firebox is used to preheat the feed entering the
reactor riser. Since feed temperature entering the reactor
riser is used as one of the manipulated variables, the firebox
constraints limit the value of this manipulated variable. The
firebox constraints are upper limits on firebox fuel flow rate
and firebox temperature. Although, both the firebox
constraints can be handled in the similar manner, it was
observed that the firebox temperature hits its constraint
before the firebox fuel flow rate reaches its maximum value.
5.2.1.1. Approximate Model for Firebox

The approximate models for the firebox were developed on
the principles of heat transfer. Due to relative fast
dynamics of the firebox, both models developed were steady-
state models.
Rate at which heat is given by firebox fuel =
Rate at which heat is transferred to feed +
Rate of heat loss
(5.1)
where
125
( T3 - Tl) - ( T3 - T2 }
,
ln ( T3- Tl) (5.2)
T3-T2
a = firebox heat loss parameter, BTU/SCF/op,

F5 = firebox fuel flow rate, SCF/s,
ABfu = heat of combustion of firebox fuel, BTU/SCF,
T1 = temperature of fresh feed entering firebox, °F,
T2 = temperature of feed leaving firebox, °F,

T3 = firebox temperature, °F,
UAf = firebox overall heat transfer co-efficient,
Therefore,
( 5. 3)
(5.4)
The parameters a 1 and a 2 are evaluated by nonlinear regression

using steady-state data obtained from the process simulator.
This steady-state data is tabulated in Appendix C (Table C.l).
The heat given to the feed by the firebox fuel must be same as
increase in sensible heat of the feed. Therefore,
(5.5)
where
Foil • feed flow rate, lb/s,
Cpoil = specific heat of feed, BTU/lb/°F.
126
(5.6)
(5.7)
The parameter a3 is estimated in the same way in which

parameters a 1 and a 2 are estimated. These parameters were
assumed to be constant throughout the control studies. The
values of these parameters are as follows:

= 0.02462 SCF/°F/s,
1. 536 X 10-4 °F- 1 1
= 25.0 lb/s.

Reactor Riser
The steady-state riser energy balance is used as an
approximate model for the reactor riser. It is explained in

detail in section 4.2.1 of Chapter 4. The on-line
parameterization of this model is explained in section 4.3.1.
5.2.1.3. Calculation of Feed

Rate for Firebox Constraints
When either of the firebox constraint is hit, the feed

temperature entering reactor riser can not be used as a
manipulated variable. Therefore, the steady-state riser

temperature target calculated using the steady-state GMC
control law is substituted in the steady-state riser energy
balance and new feed rate to the process is calculated. Both
127
the firebox constraints limit the feed rate which the system
can process. The safety buffer of 20°F is kept in the firebox

temperature constraint value and a buffer of 0.5 SCF/s is kept
in the firebox fuel flow rate constraint value.
5.3. Air Flow Rate and Catalyst

Circulation Rate Constraints
5.3.1. Approximate Models

5.3.1.1. Estimation of Carbon Concentration
on Regenerated Catalyst
The carbon concentration on regenerated catalyst is

estimated from the flue gas oxygen concentration and the
regenerator temperature using an empirical expression given by

Cutler et al. (1992).
(5.8)
where
Crgc = carbon concentration on regenerated catalyst,
Treg = regenerator temperature, K,

0 •
2 9
• stack gas oxygen concentration,
LM(x1,x2) =log-mean function= (x1-x2)/ln(x1/x2),
ln(.) =natural logarithm function,

- carbon on regenerated catalyst, reference
conditions (0.001082769),
o2air = oxygen concentration in air, 0.21,
128
0 2 r•f = stack gas oxygen concentration, reference
(0.01847),
Tregre£ = regenerator_temperature, reference (944.5 K),

EIR = activation energy I ideal gas law constant
(14991 K).
The adjustable parameters were estimated by regressing the
steady-state data given in Appendix C (Table C.2).

Concentration on Spent Catalyst
Once carbon concentration on regenerated catalyst is
inferred, carbon concentration on spent catalyst can be

estimated using an empirical expression given by Lee et al.
(1985).
(5.9)
where
kc = constant, 0.36855,
EcfiR= activation energy of coke formation reaction I
ideal gas law constant, 11679 °F,
0
Tr = reactor temperature, F.
The values of kc and EcfiR are evaluated by nonlinear
regression using steady-state data tabulated in Appendix C

(Table C.2).
129
Energy Balance
The steady-state energy balance is written for the

regenerator as follows:
(5.10)
where
cp. = specific heat of catalyst,
cpair = specific heat of

0
a~r,
Fs - catalyst circulation rate,
Fair = air flow rate to the regenerator,
Hcoke = heat of combustion of coke,
Tair = temperature of air entering regenerator,
Tr = reactor temperature,
Treg = regenerator temperature.
5.3.1.4. Reactor Steady-State

Energy Balance
The steady-state energy balance is written for the

reactor as follows:
(5.11)
where
F0 n = gas oil flow rate,
c~il = specific heat of gas oil,
130
Conv = conversion of gas oil,
Tfeect = temperature of feed entering riser,
ABcrack= endothermic heat of gas oil cracking.
The conversion of gas oil is estimated from measured variables
using following empirical equation:
Conv =
=- (5.12)
This equation is obtained by integrating the differential

material balance on gas oil across the length of reactor
riser. The temperature is assumed to be constant along the
5.3.1.5. Overall Steady-State

Energy Balance
The steady-state energy balance for the reactor and the

regenerator can be combined to get an overall energy balance.
(5.13)
The models for calculating the carbon concentrations on
regenerated and spent catalyst and for gas oil conversion can
be substituted in above equation.
131
Oxygen Balance
The steady-state oxygen balance can be obtained for the

regenerator by equating the drop in the number of moles of
oxygen in the air with the number of moles consumed in the
coke combustion reaction.
(5.14)
The constant C1 in the above equation is a conversion factor.

While calculating the value of C1 , the composition of coke is
assumed to be CH and it is assumed that the all the carbon in
the coke is converted into carbon dioxide which is an
reasonable assumption since the regenerator is operated in the
total combustion mode. The value of C1 was calculated by
material balance and was found to be 2.78. Both the air flow
rate and catalyst circulation rate in above equation have
units lb/s.
5.3.2. Parameterization of
Approximate Models
As discussed in the previous chapter, the approximate

models are parameterized using some adjustable parameters in
those models to keep them "true" to the process. The
approximate models used for constraint control are

parameterized as discussed below.
The parameterization of reactor riser steady-state energy
balance is explained in section 4 • 3. 1 in Chapter 4 • The
132
empirical expressions for the carbon concentration on the
spent and regenerated catalyst were combined to get an

expression for the steady-state rate of coke production or
coke burning. The pre-exponential in this equation was used

as an adjustable parameter to remove the process/model
mismatch. The details are given below.
The empirical expression for the carbon concentration on

regenerated catalyst ( Crqc) can be simplified as follows to
reduce the number of adjustable pa1·ameters.
K1 E 1 )]
Crgc = T.M(O. 21, 0 ) exp [ R ( Trt<UT ' (5.15)
2 •fl -~
where
(5.16)
This expression for Crgc can be substituted in expression for
carbon concentration on spent catalyst ( c.c) to get the
following expression:
------~K--------expc-E;fc ~ >>,
exp ( E ( _1_) ) r (5.17)
[ R Treg ] o.4
LM ( 0 • 21 , 0 2 )
~
where
(5.18)
S U bstl.'tutJ.'ng (C •c -c rgc ) J.'n regenerator steady-state oxygen
133
balance and solving to get the air flow rate,
_ _ _ _ exp ( _ Ecf ( ...!_) )

_ _ _ _K
(0.21-0:~ ) exp ( E ( _1_ ) ) R Tr
-rl
[ R Treg
..
LM( 0 . 21 , 0 2 )
] o.4
(5.19)
The constant K in the above equation is used as an adjustable
parameter. The parameter K is adjusted using a steady-state
IMPOL method discussed earlier so as to drive out the error in
air flow rate value to zero.
Once the expressions for carbon concentration on spent
and regenerated catalyst are parameterized, the next step is
to parameterize the steady-state energy balance equations.
Rearranging the overall energy balance equation,
= Fair cP.1r ( Tair- Treg> + FB (esc- crgc> Hcoke

(5.20)
[ -CPo.u ( Tfeed- Tr) + Conv~Hcrack]
The parameter Hcoke in the above equation is used as an

adjustable parameter. The value of Hcoke is adjusted using
steady-state IMPOL method, so as to reduce the error between
the predicted value of gas oil flow rate and the actual value
of gas oil flow rate to zero.
5.3.3. Calculation of Feed Rate

When either of these two constraints is hit, the feed
rate to the system is calculated using steady-state version of
GMC control law. When the regenerator air flow rate reaches
134
its saturation, the steady-state GMC control law is used to
calculate the flue gas oxygen concentration target. This
value is substituted in above models to calculate the feed
rate which the system can process while keeping the flue gas
oxygen concentration at its setpoint. When the catalyst
circulation rate reaches its limit, the steady-state GMC
control law is used to calculate the regenerator temperature
target. This value is substituted in the controller models
discussed above to calcuiate the feed rate which the system
can process while keeping the regenerator temperature at its
setpoint.
At every control interval, the feed rate which the system
can process is calculated using the constraint controllers.
The constraint for which the feed rate which the system can
process is minimum, is the operative constraint at that time.
Thus the objective of maximizing the feed rate is achieved
without violating any of the constraint.
135
CHAPTER 6
FCCU OPTIMIZATION
Fluid Catalytic Cracking (FCC) units are well recognized
as one of the most important units of a refinery. Their

importance derives from large product to feed stock upgrade,
large throughput (FCC units generally process 40-60% of the
refinery's products) , and their feed stock handling
flexibility. Thus, special emphasis is required to make sure
that the FCC unit is operated under the most profitable

conditions. The optimization of FCC unit is an involved and
complex task. Large number of dependent and independent
variables with their respective interactions make the

rationalization of the operation complicated.
6.1. Development of Objective Function

In order to formulate an optimization of FCC unit, it is
necessary to define an objective function. The objective in
operation of FCC unit is to get maximum profitability. As far
as the operation of an unit which was already constructed is
concerned, the cost of physical equipments, expenses for

personnel etc. will remain relatively independent of the rate
of production. If the unit is operated at almost full
capacity, then operating cost (excluding the cost of raw
material) or utility costs such as fuel gas, steam and
catalysts are relatively constant. Thus, the determining
136
economic factors which govern the operation of the unit
include the price structure of the products and of raw
materials, and the market structure, e.g., gasoline demand.
Which structure, price or market, is really controlling the
operation depends on the situation confronted by the refinery.
But for the purpose of this study, it is assumed that the
operation of FCC unit is production-limited rather than
market-limited. In other words, the price structure is a
predominant factor which governs the operation of the unit.
This means that the optimum conditions are those for which the
yield of most valuable products is maximum.
The profit is defined as the sum of the flow rates of all
the product streams multiplied by their respective unit
values, with the corresponding value of the feed subtracted
from the sum. The fresh feed is converted into gasoline, gas,
cycle oil (unconverted gas oil) and coke. The coke has zero
unit value. The economic objective function is derived as
follows:
(6.1)
where
pgr = gross profit,
Fgl = gasoline flow rate,

pgl = price of gasoline,
Fugo = unconverted gas oil flow rate,
pgo = price of gas oil,
Fg• - gas flow rate,

137
P9 • = price of gas,
F 90 = gas oil flow rate.

The gasoline flow rate is calculated from the yield of
gasoline (y91 ) as follows:
(6.2)
The flow rate of cycle oil which is unconverted gas oil, is

calculated from the weight fraction of gas oil (Yt) in the
vapor leaving the reactor riser as follows:
(6.3)
The rate of production of coke in the reactor riser can be

calculated from the carbon concentrations on the spent and
regenerated catalyst as follows:
(6.4)
where
F. = catalyst circulation rate,
c.c = carbon concentration on spent catalyst,
Crgc = carbon concentration on regenerated catalyst.
Assuming that the rest of the gas oil goes to gas completely,
from the material balance of the products, the flow rate of

gas is calculated as follows:
(6.5)
138
6.2. Optimization Models and
Parameterization
6.2.1. Gasoline Yield Model

To develop a steady-state gasoline yield model, the
spatial differential material balance on gas oil and gasoline
across the length of the reactor riser is integrated to get a
closed form solution. The temperature is assumed to be
constant along the length of the riser in order to obtain a
closed form solution. The differential material balance on
gas oil can be written as follows:
(6.6)
The differential material balance on gasoline can be written

as follows:
(-..!2..)
-R 0 tC [COR] ~ y e .RTr e (-CI tcz) (6.7)
2 0 gl I
where
COR = catalyst to oil ratio, lb/lb,
E£ = activation energy of gas oil cracking,
43700 BTU/lbmol,
Eg = activation energy of gasoline cracking,
48500 BTU/lbmol,
5 1
R1 0 • reaction rate constant, 9. 6 x 10 s- ,
139
R1 .
= reac t 1on ra t e cons t an t , 7 • 2 x 10 5 s -1 ,
1
Ro =reaction rate constant, 4.224 x 10 5 s- 1 ,

2
R = universal gas constant, 1.9872 BTU/lb0 R,

= catalyst residence time in riser, seconds,
= riser temperature,~,
= weight fraction of gas oil in vapor in riser,

= weight fraction of gasoline in vapor in riser,
a =catalyst activity decay constant, 0.1/sec,
= catalyst decay function.
Dividing equation 2 by equation 1,
(6.8)
where
(6.9)
(6.10)
Therefore, the differential equation becomes,
(6.11)
using variation of parameter method (Kreyszig, 1990),
140
( _ r2)
= e Yr • (6.12)
(6.13)
(6.14)
Using Taylor series expansion and taking the first four terms,
(6.15)
(6.16)
Then the solution of above differential equation is,
where C is a constant of integration.

Using initial condition, @yt=1, Y91 =0
ri r:
C =1r [ l - - - - 1 (6.18)
2 12
Therefore, the solution of above differential equation is,
141
(- Ez)
Yg1
= e Y~
(6.19)
This expression can be used to calculate the yield of gasoline

in the riser. This expression requires the value of reactor
riser temperature which was assumed to be constant along the
The yield of gasoline is calculated rigorously by integrating
the above mentioned differential material balance equations
and energy balance equation along the length of the reactor
riser. This value is compared with the gasoline yield
calculated by substituting the riser outlet temperature in the
above expression for different inlet conditions. Some offset
was observed in the two values due to the constant riser
temperature assumption. Therefore, average of riser inlet and
outlet temperature is used in the above expression. It was
found that when the riser temperature is calculated using
following expression, the predicted value of gasoline yield
matches that of process value. This was tested for different
inlet conditions.
(6.20)
Thus the yield of gasoline in the riser can be calculated from

the gas oil conversion value using above expression.
142
6.2.2. Carbon Concentration on Catalyst
The steady-state models for calculating the carbon
concentration on spent and regenerated catalyst are the same
models used for constraint control. These models and their
parameterization is discussed in the previous chapter.
6.2.3. Regenerator Oxygen Balance

The steady-state version of oxygen concentration
approximate model used for the control purpose is used here.
It is discussed in Chapter 4, section 4.2.2.1. The
parameterization of this approximate model is also discussed
before.
6.2.4. Steady-State Energy Balances

The steady-state reactor, regenerator and overall energy
balances are required for the optimization. These models and
their parameterization is discussed in the previous chapter.
6.2.5. Calculation of
Manipulated Variables
The feed temperature to the reactor riser, catalyst
circulation rate and air flow rate to the regenerator are
calculated by solving the steady-state energy and material
balances. The values of reactor temperature, regenerator
temperature, flue gas oxygen concentration and gas oil flow
rate are assumed and following equations are solved to get the
feed temperature entering reactor riser and hence the firebox
fuel flow rate, catalyst circulation rate and the regenerator
143
air flow rate.
Reactor Energy Balance:
(6.21)
Regenerator Energy Balance:
(6.22)
Regenerator Oxygen Balance:
(6.23)
Thus the feed temperature entering reactor riser, catalyst
circulation rate and air flow rate entering regenerator is
calculated.
6.3. Optimization Procedure
The independent variables used for the optimization are

reactor temperature, regenerator temperature, flue gas oxygen
concentration and feed rate to the system. The optimization
is done to calculate the values of these variables in order to

maximize the profit without violating any constraint.
Therefore, knowing values of these independent variables we

should be able to calculate the manipulated variables
required, e.g., feed pre-heater fuel flow rate, catalyst
circulation rate, air flow rate to the regenerator; and the

gas oil conversion and the yield of gasoline in the reactor
144
riser. Knowing gas oil conversion and gasoline yield, the
profit can be calculated as explained above.
The flue gas oxygen concentration should not drop below
1.5%, since that will violate one of the regenerator
constraints. To reduce the number of independent variables,
the flue gas oxygen concentration is fixed at 1. 6%. This
value is selected because, excess oxygen in the flue gas is
economically wasteful since energy is expended to blow air
into the regenerator. This simplification leaves three
variables to be optimized, which are reactor temperature,
regenerator temperature and the gas oil feed rate.
To further reduce the dimensionality of the problem, the
optimization is done in two cycles. The inner cycle searches
for optimum value of reactor temperature and regenerator
temperature for a given value of gas oil feed rate. The
Nelder-Mead optimization method (Riggs, 1988) is used for this
purpose. Since this is a unconstrained optimization routine,
the constraints were incorporated by using penalty functions.
The outer cycle optimizes the feed rate which the unit can
process. Since this is a unidimensional search, modified
Fibonacci method (Riggs, 1988) is used here. Thus after a
successful completion of outer cycle, the optimizer will give
the optimum value of feed rate which the unit can process and
the corresponding optimum values of reactor and regenerator
temperatures, without violating any constraint.
145
CHAPTER 7
RESULTS AND DISCUSSIONS
The Model IV FCC dynamic simulator was used to get the
unconstrained and constraint control results. The

unconstrained control results were obtained using nonlinear
process model based controller and a proportional-integral

controller. Both the controllers were tested for a set of
disturbances.
7.1. Proportional-Integral Controller
The proportional-integral (PI) controller was developed

to compare the unconstrained control responses with the
nonlinear process model based controller. The control
variables used are listed below:
1. Reactor-Riser temperature,
2. Regenerator temperature,
3. Flue gas oxygen concentration. /
The manipulated variables used are listed below:
1. Riser feed temperature,
2. Reactor/Regenerator pressure differential,
3. Regenerator air flow rate.
The riser feed temperature has a direct and dominant effect on

the reactor-riser temperature than on other control variables.
Therefore, the reactor-riser temperature was controlled by
manipulating the riser feed temperature. The relative gain
146
analysis was performed to find the proper pairings for the
control of regenerator temperature and flue gas oxygen
concentration. The relative gain was found to be 1.72 for the
flue gas oxygen concentration and the regenerator air flow
~
rate pairing. Therefore, following pairings were selected for

the PI controllers.
1. Reactor-Riser temperature and Riser feed temperature,
2. Regenerator temperature and Reactor/Regenerator pressure
differential,
3. Flue gas oxygen concentration and Regenerator air flow
rate.
Basic regulatory PI controllers were used to get the
desired riser feed temperature, regenerator air flow rate and
the reactor/regenerator pressure differential. The regulatory
PI controllers for riser temperature and the regenerator air
flow rate are discussed in Chapter 4.
The reactor/regenerator pressure differential was
controlled by manipulating the valve (V14 ) on the regenerator
flue gas line. The manipulation of this valve changes the
regenerator pressure. The reactor pressure is maintained at
its setpoint by manipulating the wet gas compressor suction
valve.
7.2. Tuning of Control Loops

There is a hierarchy of control for both the nonlinear
PMBC and PI control strategies. The reactor-riser
147
temperature, regenerator temperature and the flue gas oxygen
concentration control loops at the upper level and the basic
regulatory PI controllers for riser feed temperature, reactor
pressure, catalyst circulation rate, reactor/regenerator
pressure differential and the regenerator air flow rate at the
lower level. The basic regulatory PI controllers were tuned
first. Each of these PI control loops was tuned using
setpoint changes in the controlled variables. The initial
tuning was done to minimize the integral of square of error
(ISE), defined as
(7.1)
where
e = error between the controlled variable and its
setpoint.
The minimum ISE criterion puts more weight on large errors,
which occur at the beginning of the response, and less weight
on the smaller errors, which happen towards the end of the
response. In trying to reduce the initial error, the minimum
ISE criterion results in high controller gains and oscillatory
responses with the error oscillatory about zero for a
relatively long time. Therefore, the fine tuning was done by
visual inspection of control result to get a fast response
without too much overshoot and without too much oscillation
about the new setpoints. The velocity algorithm was used for
the PI controllers given by the following equation.
148
(7.2)
where
Am = change in manipulated variable,
en • error in the control variable when control action

is taken,
en-1 = error in the control variable at the time of
previous control action,
Kc • proportional gain,
~1 = integral time constant.
The tuning constants for each of the basic regulatory PI
controllers are listed in Table 7.1.
For upper level PI controllers, initial tuning parameters
were calculated using Cohen-Coon tuning rules. These tuning
parameters were updated further as discussed above. Table 7. 2
lists the tuning parameters for these control loops.
For nonlinear PMBC controllers, initial tuning parameters
were set by trial and error and adjustments were made to get
minimum ISE. The tuning parameter for the integral term was
set to zero for all the control loops, since the frequency of
on-line model parameterization is greater than the frequency
of control action. The tuning parameters for nonlinear PMBC
loops are listed in Table 7.3. The tuning constants for the
constraint controllers are given in Table 7.4.
149
7.3. Unconstrained Control Results
The nonlinear process model based controller (PMBC) and
the proportional-integral (PI) controller were tested for a
set of disturbances. The performance of both the controllers
is discussed below.
The responses shown by Figures 7.1 to 7.3 compare the

performance of PI and PMBC controllers for Disturbance 1.
Disturbance 1 includes positive and negative step changes in
coke formation rate constant. This is a primary disturbance
affecting the operation of FCCU, which results due to the
different coking characteristics of feed used. This
disturbance is modeled in the simulator using a coking
formation rate constant. Disturbance 1 include positive 2%
step change in the base value of coke formation rate constant
at 300 seconds. This value is brought back to base case at
20000 seconds. A negative 2% step change is given in this
value at 40000 seconds to simulate the feed having less coking
characteristics. The reactor temperature response to this
disturbance is shown in Figure 7.1. The PMBC response is
faster without much deviation from the setpoint throughout the
length of the disturbance. For PI controller, the reactor
temperature deviates by 3°F from the setpoint when the step
changes are given and takes about 7000 seconds to return to
the setpoint. For negative step change at 40000 seconds, the
PI controller response shows small oscillations while the PMBC
response is without any oscillations. The regenerator
150
temperature response to disturbance is shown in Figure 7.2.
For positive step change at 300 seconds, the regenerator
temperature shows a deviation of 3°F above the setpoint for
both the controllers. The regenerator temperature comes back

to setpoint in 4000 seconds for PMBC controller and in 9000
seconds for PI controller. For a negative step change at

20000 seconds and 40000 seconds, the deviation from the
setpoint is more for PI controller than for the PMBC
controller. Also the PI controller response is more sluggish

than the PMBC response. The regenerator temperature open-loop
response to this disturbance is nonlinear with higher steady-
state gain for negative step change. As seen in this result,

the PMBC controller understands the nonlinearity of the
process while PI does not. The flue gas oxygen concentration

response to disturbance 1 is shown in Figure 7.3. The PMBC
response is faster than the PI control response. Also, the
deviation from the setpoint in the flue gas oxygen

concentration is more for the PI controller than for the PMBC
controller.
The responses shown in Figures 7.4 to 7.6 compare the
Disturbance 2 includes the positive and negative step changes

in the fresh feed temperature entering the feed pre-heater.
A positive 20°F step change is given in the fresh feed
temperature at 300 seconds. The fresh feed temperature is
brought back to the base case at 20000 seconds and a negative
151
20°F step change is given at 40000 seconds. The reactor
temperature response to this disturbance is shown in Figure
7.4. As seen in the figure, the reactor temperature moves by
approximately 0.5°F from the setpoint but immediately returns
back to the set point. The response of both the controllers is
almost identical. The regenerator temperature response is
shown in Figure 7. 5. The deviation in the regenerator
temperature is slightly higher for the PI controller than for
the PMBC controller. The PI controller takes approximately
7000 seconds to return to the setpoint. The response of both
the controllers is very good for both the positive and
negative disturbances. This is because the step change in the
fresh feed temperature is taken care by the riser feed
temperature controller by manipulating the fuel flow rate to
the feed pre-heater. Thus the feed temperature entering the
reactor riser is not affected much and thus have a very small
effect on the reactor-riser temperature. Since the reactor
temperature is controlled well, the effect of this disturbance
on the operation of regenerator is very small. The response
of flue gas oxygen concentration is shown in Figure 7.6.
The responses shown in Figures 7.7 to 7.9 compare the
Disturbance 3 includes positive and negative ramp changes in
the ambient air temperature. The changes in the ambient air
temperature affects the capacity of the regenerator air
blowers, thus affecting the operation of FCCU. Disturbance 3
152
includes 30°F positive ramp change in the ambient air
temperature at 300 seconds. The ramp is applied over a 60
minutes period. In a real FCCU, this level of disturbance
would occur over 6 to 8 hours, as part of normal daily cycles
of daily ambient conditions. The 30°F negative ramp change is
applied starting at 20000 seconds to bring the ambient air
temperature back to the base case value. A 30°F negative ramp
change is applied again starting at 40000 seconds to bring the
ambient air temperature below the base case value. The
reactor temperature response to this disturbance is shown in
Figure 7. 7. The reactor temperature does not deviate from the
setpoint for both the positive and negative ramp changes. The
PI controller response shows very small oscillations in the
reactor temperature at the end of negative ramp starting at
40000 seconds. The PMBC controller response is smooth. The
regenerator temperature response is shown in Figure 7.8. The
PI controller response drifts approximately 1.5°F away from
the setpoint over the time period in which the ramp is
applied. The PMBC controller response is tighter than the PI
controller response and the regenerator temperature stays at
the setpoint throughout the length of the disturbance. The
flue gas oxygen concentration response is shown in Figure 7.9.
The PMBC controller response is slightly better than the PI
controller response. The flue gas oxygen concentration show
slightly higher deviation than the PMBC controller for the
negative ramp change in the ambient air temperature.
153
7.4. Constraint Control Results
The constraint controller was tested for the same
disturbances for which the unconstrained control strategy was

tested· Following is the discussion of the constraint control
results.
The constraint control responses to the coking

characteristics disturbance (Disturbance 1) are shown in
Figures 7.10 to 7.15. This disturbance includes positive 2%

step change in the coke formation rate constant at 300
seconds. The coke formation rate constant is brought back to
the base case value at 20000 seconds and a negative 2% step

change was given at 40000 seconds. The maximum air flow rate
which the regenerator air blowers can provide was fixed at 78

lb/s for this run. When a positive step change is given in
the coke formation rate constant, the air flow rate to the
regenerator starts increasing to burn the increasing amount of

coke coming from the reactor. As seen in Figure 7.11, the
regenerator air flow rate reaches its saturation limit of 78

lb/s at approximately 10000 seconds. Therefore, the air flow
rate constraint becomes operative and the feed rate to the
process is manipulated by the air flow rate constraint

controller. When the coke formation rate constant is lowered
at 20000 seconds, the amount of coke to be burnt in the
regenerator decreases and thus the air flow rate required to

keep the flue gas oxygen concentration at its setpoint also
drops. Since the decrease in the coking characteristics of
154
the feed drops the reactor temperature, the fuel flow rate to
the feed pre-heater required to keep the reactor temperature
at its set point increases. Due to this, at approximately
21000 seconds, the firebox constraint becomes operative and
the air flow rate constraint no longer remains operative.
Therefore, after 21000 seconds, the feed rate to the process
is governed by the firebox constraint. As seen in Figure
7.11, the air flow rate to the regenerator drops below its
saturation limit, since this constraint is no more operative.
As seen in Figure 7.12, the firebox temperature remains at its
constraint value of 1680°F. The firebox constraint remains
operative till the end of the simulation. The responses of
control variables are shown in Figures 7.13 to 7.15.
The constraint control responses to the fresh feed
temperature disturbance (Disturbance 2) are shown in Figures
7.16 to 7.19. The disturbance 2 includes positive 20°F step
change in the fresh feed temperature at 300 seconds. The
fresh feed temperature is brought back to the base case value
at 20000 seconds and a negative 20°F step change is given in
the fresh feed temperature at 40000 seconds. The firebox
constraint becomes operative at approximately 40000 seconds
when a negative 20°F step change is given in the fresh feed
temperature. The decrease in the fresh feed temperature
forces the fuel flow rate to the firebox to increase, thus
hitting the firebox constraint. Thus starting at 40000
seconds, the feed rate to the process is manipulated by the
155
firebox constraint controller. The feed rate response is
shown in Figure 7 .19. The control variable responses are
shown in Figures 7.16 to 7.18.
The constraint control responses to the ambient air
temperature disturbance (Disturbance 3) are shown in Figures
7.20 to 7.24. The Disturbance 3 includes positive 30°F ramp
change in the ambient air temperature at 300 seconds. The
ambient air temperature is brought back to the base case value
at 20000 seconds and a negative 30°F ramp change is given in
the ambient air temperature at 40000 seconds. The firebox
constraint becomes operative at approximately 2000 seconds and
the feed rate to the process is adjusted by the firebox
constraint controller. The feed rate response is shown in
Figure 7. 24. The control variable responses are shown in
Figures 7.20 to 7.23.
7.5. Optimization Studies

The steady-state optimization models were used to perform
the parametric studies. The effect of riser temperature on
gas oil conversion and the gasoline yield was studied. Also,
the effect of catalyst carbon concentration on the yield of
gasoline was studied. The Figure 7. 24 shows the plot of
gasoline yield versus gas oil conversion. As seen in this
figure, the gasoline yield initially increases with increase
in gas oil conversion, but after certain gas oil conversion
the gasoline yield starts decreasing with increase in the
156
conversion of gas oil. The effect of riser temperature and
the catalyst carbon concentration on the yield of gasoline is

shown in Figure 7.25. For a fixed value of catalyst carbon
concentration, the gasoline yield first increases with
increase in riser temperature. But After a certain riser

temperature is crossed, the gasoline yield starts decreasing
with the increase in the riser temperature. It can be seen

that the optimum in the gasoline yield lies above a riser
temperature of 995°F for the operating range of catalyst
carbon concentration. Since the reactor-riser can not be

operated above 995°F due to the metallurgical temperature
limit, the riser should be operated at its constraint in order
to get the maximum yield of gasoline. As seen in the figure,

when the riser is operated at 995°F, increase in the carbon
concentration in the catalyst decreases the yield of gasoline.

Therefore, to get better gasoline yield, the better
regeneration of the catalyst is essential.
The optimization was done off-line to evaluate the

optimum reactor and the regenerator temperature for different
coking characteristics of the feed. Also the off-line

optimization was performed by varying the prices of products
over a possible range to find the their effect on the optimum.
The results of off-line optimization are tabulated in Tables
7.5 to 7.7.
157
Table 7.1 : Tuning Constants for Basic Regulatory PI
Controllers
Controller K1 'CI
Riser Feed Temperature 1.0 (SCF/s) /°F 333.3 seconds

Controller
Reactor Pressure 0.1 ps.L• -1 1.25 seconds

Controller
Catalyst Circulation 1. 5 X 10-4 33.3 seconds

Rate Controller ( lb/s) -l
Reactor/Regenerator
Pressure Differential 0.08 ps.L• -1 1.3 seconds
Controller
1
Regenerator Air Flow 0.05 (lb/s)- 19.0 seconds
Rate Controller
158
Table 7.2 : Tuning Constants for Upper Level PI Controllers
Controller Kc t:I
Reactor Temperature 3. 0 °F/°F 9.5 seconds

Controller
Regenerator temperature 0. 0225 psi/~ 24.7 seconds

Controller
Flue Gas Oxygen 175.0 lb/s 3.7 seconds

Concentration
159
Table 7.3 : Tuning Parameters for Nonlinear Process Model
Based Controller
GMC Control Law K1 K2
Reactor Temperature 1.2 0.00
Regenerator Temperature 0.0025 s -1 o.oo

Flue Gas 0 2 Concentration 0.004 s -1 0.00
160
Table 7.4 : Tuning Constants for Constraint Controller
Controller Kl K2
Firebox Controller 1.1 o.o

Air Flow Rate 0.15 0.0
Controller
Catalyst Circulation 3.0 o.o

Rate Controller
161
Table 7.5 : Optimization Results for P~0 • 0.45 $/gal,
P9 i = 0.52 $/g·al and P9 • = u.48 $/gal
Reactor Regenerator Profit at

Case Temperature Temperature the Optimum
6
(oF) (oF) (10 $/day)
+2% Coking 994.40 1280.25 0.0975
Base Case 994.99 1280.25 0.0993
-2% Coking 994.99 1280.25 0.1008
162
Table 7.6 : Optimization Results for P~o = 0.45 $/gal,
P91 = 0.55 $/gal and P9 • = u.48 $/gal

(oF) (oF) ( 10 6 $/day)
+2% Coking 994.40 1280.25 0.1211
Base Case 994.99 1280.25 0.1233
-2% Coking 994.99 1280.25 0.1253
163
Table 7.7 : Optimization Results for P~0 • 0.45 $/gal,
P91 = 0. 52 $/gal and P9 • = u. 60 $/gal

(oF) (oF) (10' $/day)
+2% Coking 994.43 1280.25 0.1307
Base Case 994.99 1280.25 0.1339
-2% Coking 994.99 1280.25 0.1369
164
l'z.4
0\
~
-CD
.....
1001 ·-·--
.B 997 - ·
Cd
.....
Cl)
~ 993 ·----------·-·-
(1)
E-4
..... PID
.80 989 ·------··--··
«S
~
985+-----~----~----~----~----~----~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.1 Unconstrained Control Response of Reactor

Temperature for Disturbance 1
165
~ 1285~--------------------------------------~
Ol
Q)
-
~ 1280 - -·- - - ---·---------------------------------
~ 1265 -------------·----------
0
,a)
Cd
~
PID
Q) 1260 -----------------------------------
a=
Q)
Ol
~ 1255+-----~----------------------------~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.2 · Unconstrained Control Response of Regenerator

166
dP
- 4~--------------------------------------~
~
-~ 3.5 ··-·-----·------·-------···--·------··-----··-·-·---·--···-···-·····-------·---··----- ··----·-····
+)
ctS
~
+)
3 ··-·---·---·--··--·---·-···--·--·-·-······----------·--·---··-·-·--··--·--·-·-··-······--···-········--···--·······-····-···-·--···----·
~
G) PID
u 2.5 -·--·-----···------------·-·····-·--··-··--···--··-·-·-··-···········-·-··---·-··--·-····--
~
0
0
~
G)
t1l
~
><
0 1 ---------·-····--···-·----·------ - ·-···----······----···--·-····-···---·----·········---···
m PMBC
~ 0.5 ·- ·--··-········-·---···-·-····--···-··-··-----··-·---·-------······--··························-··-········-·-··--··
0+------.------.------.------.-----~------~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.3 Unconstrained Control Response of Flue Gas

Oxygen Concentration for Disturbance 1
167
1005~=============--~__::~--------.
-
Ol
~ 1001 - ···--------------------····-·········-·-····--·--------------------------·-···
- PMBC
997- - / -----------------------------------·····-····--·-------------------
993- . " ' \ --- --·- --~-- - - __________.':__________________________ _
PID
989- --------------------······--·--:··--·--············--·-···------·····-·--------···-····-····-··-·-----------·---·-····--··
985+-----.-----.-----.-----.-----~----~--~
o 1'o 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.4· Unconstrained Control Response of Reactor

168
Ol
-
Q)
"0 1280 --··-·--····--···········-···-·-·---···---·---·--·-·-·-·--·--··--·--··-·······-··--
Q)
~
:::1
....,
ld
~
Q)
~ 1270 -· -
Q)
E-i
~ 1265 --····-··-----·---~-~!?._____________________________________________________________________ _
....,
ctS
~
Q)
~
1260 ·---·--··-·----·-···-···-·-··--·------------····--·-·-·---·····--·----···---·-···
Q)
Ol
~ 1255+-----~----~--~----~----~----~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.5 ·u nconstrained Control Response of Regene r at or

169
-§
* 4~--------------------------------------------~
3.5- ··-··--·-··---··-·-····--·--··----·- - - - -·-·-·-·--·-·-··-·--······--·--···---·····-

....4.J
~
1-4 3- ··--·--·--·---- - - - -·-··-- ---- --··-·--·······-·-···--·----·····
4.J
~
~ 2.5- ---··--···---------·-·--·--·-·- -·-·--·----··---·-·--·--·-·-···--····-----···----·--·······
~ PMBC
~ 2- -- -~-------------------T-----------····-------------
~ 1.5- ····-· ..... -····· ·-- ·-· ·-·-· ... ·-·· . -··· -, ----· -· .
0 1 - ······-~·-···-·-·-··-····-·····-·-·--····--·--·--··----~---··--··--·-··-·-···-···-·-·····-·-··-···-········-·
: PID
{!) 0. 5- -···-·-·-··-····--·-·--·-·······-·-·····-···-··--·····---- - -····-·-----········-·-··---····---····-···-··
0+------.-----.------.-----~----~-----.----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.6 Unconstrained Control Response of Flue Gas

170
1005~------------------------------------~
fz.t
tTl
1001 ···----··-··-····------···-·----------····--··-···---------
-
Q)
"0
PID
997 -_---·--------·--·-·---- --~----·-··--···--·--
993- -------···-·--------7 ·----·--··-= -··-------·-·-·----·- -
PMBC
989- ··········--··---·--··---·-··--·-··-··-·······-·-·-··--·····-·--·-··-··-·-·-·-··-····-····················---··-·····-··-···
985+-----~----~------~-----r~----~----~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.7 Unconstrained Control Response of Reactor

171
1285.---------------------------------------~
Ol
~ 1280- ·-----·--·------·········-····-···---·---·······--·--·--·-··----·-·······--···-··-·--·--------··-······--·-····-···
-~ PMBC
~ 1275- - - - -- ; r------·--···--·-----···-----··-···------····-········-····---·······
e0.. 1270- .... -............

~
········-····-····----·-···- -·········- -- ···-- - ····-·-·· . ·-··-······-···-··--··--··-·····-··
8
PID
$.4
0
1265- ············-··-··-······--···············--······--------····--------···-···-···-··-···-····--------··········--···--···--··
4-J
«J
$.4
~ 1260 .···--····--··--··················--················-···-·-··-·····----·-····-···-····-----···--·--········-···----····-····-·
~
~
Ol
~
a:t: 1255+-----~------~
~ ----~----~T------~----~----~
0 10 20 30 40 50 60 70
Time (Seconds )
(Thousands)
Figure 7.8 · Unconstrained Control Response of Regenerator

172
* 4~------------------------------------------~
-
§ 3.5- ·--····--··-··-··----·---·--·-----------------···--······-···-·---··-·-·--··
·ri
+)
cu 3- ·····-··········-·-···-···-······--······--····-··-·····-·-····-··········-····--·---····---····-··--··········-···············--·-----
~
+)
~
Q) 2.5- ············-·········-··-·-·······-········-----·--····----------····-·---- - - - - - - - -
u PID
~ 2- ····- - - ----·-·-·--------- ----;/--- - - - - - -·--···--·----·-
! 1.~-
cu
t!)
•::: _·::• ~:•• ----~-::: ~~-:~:--~=~=-~=:~: ·:~~~~~~
:: __-
PMBC
_
0. 5 ·····-····-··········--··-····-····-········-··-·-····-·······--------··-·--·---··-··-·----··-·········-----·-··--··--·
0+-----~------~----~------~----~----~~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.9 · Unconstrained Control Response of Flue Gas

173
150~------------------------------------~
140 --·--·-----·-----···-----···------····-------····--
CD
Jl ~ fV-"vy"\ A
~ 130 ·-·-----·------- · - - - · · - - - ·-· ·····--·---------··--
Q)
+)
~
~ 120 -·---···----·-··-···--·······---·-··-·---·-·-·-·----··------··-·-···------·······-··-·-··--··-··-·
"tj
(l)
(l)
rz..
11 0 ···--··---···-·---····-··--···-······--·-··--·--·-·-·-··---··-··--·-·-··-········--·----·-·-·-·-·---·········-··
100+------.-,----.------.-----,.-----.------r----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.10 Feed Rate Response with Constraint Control

for Disturbance 1
174
m eo~----------------------------------------~
..........
.Q
~
- 85 ····--·-·--·-·····------·----···-·-··--·--·-·····--------··-····-·-·-··--··--·-··--······-·
G)
....,
cs
~ 80 ··------------··-----------·--·-·-·-·---·----·-·--··---·-·-···---·····--·-·······
~
~
rz.. 75 -···-=-·---·-··--···---·-·--·-···--:::::
···-=----.,..,.1
~
·ri
<
~ 70 --·-··--····-···-··-···········-·······--···········-······-·--·-···--···-·--·-·-·-····-·-·-··-···-··-·-·······--··--
0
....,
cs
~
G) 65 ·-···----··--·---··-···-·---·-········-··--··--·--·-·--·--··--·--·---·---·-···-··--····-········-··
c:=
G)
tJl
~ 60+-----~----~----~----~----~------~--~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.11 Regenerator Air Flow Rate Response with

Constraint Control for Disturbance 1
175
1900.-------------------------------------~
~
Ol 1800 ·-·-·----··-·--···-··--·---··-··--·-···----·-····-·····-·-··-··-··-·······-·--·--··--·----------·········---··--····---·
Q)
"0
- 1700 ···-----···-·--·----···-·-····--·------···----···-·····-···--·--·-··---·--····--··---··-··-··----···-·-····------·--·--·-···-···-··---
,..
Q)
~
...., ·--·-··--·----------·--·-··-·······-·-···-··--·-·--·-------·--·-··-
,..aS
Q)
~ 1500 ·- . . ...... ····-···-·-··--··-·····-··-·-·--··········-····-···-·-·-··-·--·····--·-··-·······---··-·-···--············
Q)
8
>< 1400 ······-·--··-··-···········-··-····--·-···--···-··-·-·-·······-·············--·-·····-·-------·-·······-·········----····--···
0
..0
Q)
,.. 1300 -·········-··-···-·······-····--·····---·-··-·-··-·-·---··--·--··-··········-·--···-···----··.:··--··-··--··-·--···--··
.....
~
1200+-----~----~--~----~----~----~----~
0 10 20 30 "40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.12 . Firebox Temperature Response with Constraint

Control for Disturbance 1
176
1005~========~-=========~~~----~
g' 1001
-
"tj
G)
5-f
...,::J 997 ·-·--··-···------··-·---···-·-·--···--·--·-----·----------
as
5-f
CD ,l~.o- .l.tA IJ. 1
~ ,.,- .. ,.. y ., f ·r
a;; 993- ..·-·-····--·-······-··-·---·--·--·-·--·····..····--·--·····-··-·-·-····-·····-·--··-·-·-·--·-...... _.......
Q)
8
5-f
0
+) 989 ·--···-----·--·-···-·-··---·······-····--··-----·- ..-··--·"""'-----·-··
u
as
~
985+------,------.------T,-----,.------~-----.----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.13 Constraint Control Response of Reactor

177
Ol
~
-C1)
,...,
1280 ···-···---·-·-------·-··---·--·--···-··-···-·-----·---··--·- ·-·--·-------·-
~ 1275 --- --·---···--·- ·---·-- · - -

.,,...,
.&J
C1)
~
C1)
E-t
,0. ., 1265 -···-··-·-·-·--·-·-··--·-···--------·-------··--···-·-·---···----···-····-·····--·----·
.,
.&J
,...,
C1)
~
1260 ·-- ·--------·-·--·-·-·-·--··-····-----·--··-·--···-····--··-··-···--··-····-·····-·
C1)
Ol
~ 1255+-----~----~--~----~----~----~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.14 Constraint Control Response of Regenerator
178
4~----------------------------------------~
-
8 3 .5 ·-·--·----·-·········-·----·-········-···-·····--·····-·---···-··--····---·--·---······-·····-············-····--··---·
· ri
~
cu 3 ·····-------·-·---········-·---·----··-··----·---·······-·-··--·--··-·--··--····-·········--··········-·--··
~
~
c::
CD 2.5 ··-··----·-·----·-··--···--···-····--··-·-·········-·-····---·------··---·····-···········-·-·--·····-··-····
0
c::
0
u 2
c::
CD
O't
:>,
1.5
><
0
1 ·---------···-·-----·-··-···--······················-··-··-···-······-··---·----······-·-··-·-·-···-·-·········--······-····-······
co
cu
{.!) 0. 5 ····--····-····-·-·····-··············-···-···-··-···-··---··-·-··-···-·--···----·---·--····-·······-···-·······-·········-······--··
CD
~
,..-4
IZ-4 0
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.15 Constraint Control Response of Flue Gas

179
1005~------------------------------------~
-
0"1
1001 - -----------------·-------------------------------------------------------------------------------------------------------·
-
CD
~
997 ----------------------···-····-··-·--···-······----··-···-···-----------·--··--·····-··-·-·
..
993 11- -···-·--··-···-··-·--···--·-----·--··-·····-··-·-·-·-····-·······-···---·--·-···-···--··--·-······---·---·-··-·
989 ···--·-···----·············------··-··--···-··--··-·····----··---······-·-·-·······-·········-··-······--·····
985+-----~----~----~----~----~----~----~
0 10 20 3040 50 60 70
Time (Seconds)
(Thousands)

180
1285~------------------------------------~
0'\
~ 1280 -·······--·····-··-·-··-·-········-··-·-··-··-·······-···-·-·-·-·--·······-·--··------·-··-··········-······--···-····---·
-
Q)
~
~ 1275 --··-·---·····-·-········--···------·····-····--·-···------··--··-··-···-···-·--··-·-·-······-·-·-···-····---·········
.,
.&J
i
~
E-t
1270 tF·-····----····- --··-· --·--·· ·--·-· --·--··-··-·-----·
~
0
1265- ·····-·····-········-······--·-·······-······-·················-··--···-·····--··-··············-·-··········- -·········-····-·----·--·
.,
.&J
~
Q)
J::
·1260-················--·-··--·····-···-·····-··············-·-·-·······-···-----·-··---·-·---···········-- ···-········-··
Q)
0'\
Q)
~
1255+-----~----~----~----~----~----~--~
0 10 20 3040 50 60 70
Time (Seconds)
(Thousands)

181
-a 3.s -----------------
4~--------------------------------------~
·ri
+»
ld 3 ·---------------------------------------------------··-·-·---
J.f
+»
c::
u 2.5 ·---------··-------··-·---··---·----·--- --·----------··--·------··-
Q)
§
u 2 ·--------·------·---·-----------------------·------·-··--·------------
~
1.5- r - - · · - - -----·-----··-·--·-·-·--·--·-·····-···---·· ~ ··------·----·-- -·· ·----·

Q)
0\
>,
~
0
1 ----··------·-·----·---···-··-·----·--····-----···--·-·---------·--------·---·------··----·
Ill
ld
(!) 0.5 -----------------------------·-··--···--·····----··········-·-···---------·----·-·-------·-----·
Q)
~
~
rz. 0+------.------,------.------~.,------,~----~----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)

182
150~------------------------------------~
1404-----·--------------··------·-----------···-----
-CD
......
.Q
~ 130~--------··-·----·-·----- .=. ....=.=
.. _= .... ····=·-·=···----··===----- .
Q)
""~«S 120 ···----··-··········----···-·--··-----··-·-·········-·--·--··-·-····-------······--·-·····--·······---···-··--

"tj
Q)
Q)
~
110 ····-------·--·-·-·-······-·-··-······---···-·--·---------··-----······--······---·
100+-----~------~----~----~------~-----T
~----~
0 10 . 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.19 Feed Rate Response with Constraint

183
1005~------------------------------------~
-
0'
cu 1001 ·- ····--·--·-----·--·--···-···----····--··--·-···-·-··--·-··-··---·····-···--···---··-···
-
"tj
997 --·-·-··-·---· -····- - ·-··---···-··-·····-·-·-·--·--------··----·-----···-·-··
.....
"'I' . ..,
993- ·-·-·-··---····--·-······-··-··-··-····-···-··-····--···---·-----------·---··---·····-······--·
989- ···---·-·······---··········-··----···--·--·····-······---··--·---·-·-··--·-······--···----··----··
985+-----~----~----~----~----~--~~--~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)

184
~ 1285.-------------------------------------~
0'1
~ 1280 ·······--·-·····-·········-··-···---············--·-····-··-···---····-···-····----··--·-···--·--·-·----·
-Q)
~
~
,&..)
1275 ·-··-·--·-···········-···········-··-·-··-·-··-·-··-···-·---··--·-·-··-··--··----·--····--··-·-·---·······-·-·-
cu
t
~
E-4
1270 t;P ------------· -·---·-··-------=. ·-·-·--- -···---···-·--
~ 1265 ····-········-·--··-·-······---·-··-···················-·-··----·-···-·-··-··---····-·····--·-···-·-·--···-···-·
.&J
cu
~
Q) 1260 ···-·--- -·-·--:-··-····-····----·---·······-··-----···-·------·-·-;···-··-·-·----····-·---···
aQ)
0'1
Q)
p:: 1255+-----~----~------T~----~~------~~------~--~
0 10 20 30 40 50 60 . 70
Time (Seconds)
(Thousands)

185
* 4.---------------------------------------------------~
-
§ 3. 5 ··-··-······--··-·-·····-·--····-··---·-·-·--·-·--··----··-···-·-··-··-··----···---········-··-·-···-··--····---
.....
+'
coJ-1 3 ·-·--··-······-----·---·-··-·--····--········-········-···-·-·-·-·--···········--···-·-····················-····--·----·-·-···
+'
s::
~ 2.5 ··-----·--···--····---·-··-··-----·--· - -
s::
0
t> 2 ··-·-·-··--·····-·-····-·······-··--········-····--·-·······-·-·······---············-····-·---···-··-···--·-···-············-····--------·-···
1 ··---·····--·-···-···-···---·····-····-··-·····-····················-·······-·············--·-··--·-····-······-·-·········--·-·-··········-····-················
ell
Qj
t!) 0. 5 ·-········-··-·····-··-··--··--·-·-·-··-··--·······-···-·-···-······-···-··--·-··-·---·-·-··-·-··-·-·--····-·-····-···········-·-··
Q1
~
=='
0+-------~----~------~------~------~----~------~
r:...
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)

186
150.---------------------------------------~
140 ····-····--···------·-·---··-·- ·-·-·----·-·--··-···-·- -·-···--···--·--····

m
'
~ 130 ~--·--·--------------·----~~·-------·---
....,0)
co
J..f 120 --···-···-·-····-··----·-·----·--············-···········-·-··-····-······-····-·······-------·-·····-·-·······-·--····--··--·
"0
Q)
Q)
~
11 0 ···-·-·-·····-··---······--···--···--·····-·········--·-···-··---··-·--·----···-··-·-··----··--···-·····
100+------.------~----~----~------~----~T----~
0 10 20 30 40 50 60 70
Time (Seconds)
(Thousands)
Figure 7.23 Feed Rate Response with Constraint

187
48~--------------------------------------~
dP 4 7 --------------------------------------------------------------------···---···-·······-··-····-···---···-·-··-·--··-·-·-··--·········-
~
-
~
"'d 46 -··-················-·-·-·······-·····-··-···-········ ··-·········-··---··--·······-·-··········--·-· ·--····--·--··-··-·-·····-··-·····-···

...-t
Q)
.....
~
Q) 45 ··········--···-·-····--··-·········· ·-············-···········-·····-·······--·····--·-··-······--·-·················- ··-·······---················--·--······

.....~
...-t
0
G)
~ 44 ·················-······························-·····-···········-······-··-········--··-··-·-········-··-····-·····-···-····-· -···············--····----·····
43+-------~------~----~------~------~------~
60 65 70 80 75 85 90
Gas Oil Conversion (Wt %)
Figure 7.24 Plot of Gasoline Yield versus Gas

Oil Conversion
188
,... Constraint Boundary
'1j 0. 4 7 ···-·--··-·-·····-·-··-·····-··- -·-·-···----·····---·--···----··--·---·-·······-··-·- - - -·-···

r-4
Cl)
·ri
~
Cl)
c::
·ri
r-4 Increasing
g Coke Cone.
ld
~ . .
0.45 ··----·-·-··· ··- ·-· .... ····- ···-·-·····-····-·-··-····-··-·-···--····-··-··-···-··-·· -·· ....... ····--····-······- ······-·····
995 1030 1065 1100

Riser Temperature (deg F)
Figure 7.25 Effect of Catalyst Coke Concentraion

on Gasoline Yield
189
CHAPTER 8
CONCLUSIONS AND RECOMMENDATIONS
The main purpose of this work was to apply nonlinear

process model based control to the Amoco/Lehigh University
Model IV Fluidized Catalytic Cracker challenge problem. A
dynamic simulator written in ACSL (Advanced Continuous
Simulation Language) which was provided by the authors of this
challenge problem was converted to FORTRAN and was bench
marked against the provided open-loop responses. The riser

model from the original simulator was replaced by a plug flow
reactor model with a simple yield model. This modification
made it possible to apply a meaningful on-line optimization
analysis as well.
The control strategy was first tested for an

unconstrained case. The nonlinear process model based
controllers were developed for the control of reactor
temperature, regenerator temperature and the flue gas oxygen

concentration. The dynamic macroscopic oxygen and energy
balance in the regenerator while the steady-state energy
balance in the reactor were used as nonlinear models for
control. The approximate models were continuously
parameterized using IMPOL method. The Generic Model Control

(GMC) law was used for the nonlinear PMBC controller. The
unconstrained controllers were tested for a set of
disturbances. The unconstrained control results were also
190
obtained using a PI control strategy. The PMBC control
results were found to be better than the PI control results.
After getting the unconstrained control results, the
constraint controllers were developed for constraints on feed
pre-heater, air flow rate to the regenerator and the catalyst
circulation rate. The feed rate to the system is used as a
manipulated variable based upon maximizing process throughput
without violating any of the constraints. The constraint
controller models were developed relating the feed rate to the
process variables. The constraint controllers were tested for
a set of disturbances.
The last part of this project was to do the optimization
studies. The steady-state controller models were used for the
optimization. In addition, the steady-state models predicting
the gas oil conversion and the gasoline yield were developed.
The optimization was done off-line for a range of prices of
gasoline and wet gas. Also parametric studies were done with
the optimization models.
8.1. Recommendations
Due to time "constraint," various aspects of this project
are still remained to be explored. The recommendations for
further work on this project are listed below:
1. In this work, the catalytic cracking was modeled using
Weekman's "three lump" model. This Model combines the
chemical species into three pseudo-components: gas oil,
191
gasoline and light gases/carbon. Although this yield
model allows us to do the economic optimization, this is

not a detailed model. To perform a realistic
optimization, a detailed yield model is required which
can predict the composition of various products formed in

the gas oil cracking.
2. To make the simulator more realistic, noise should be

added in the measured variables and drift should be added
to the process parameters.
3. In this work, the flue gas oxygen concentration and the

regenerator temperature were used as control variables
for the regenerator. As discussed in Chapter 5, there is
an alternate set of control objective present for the

regenerator. It involves control of flue gas carbon
monoxide concentration control instead of flue gas oxygen

concentration. The control of carbon monoxide
concentration is more difficult than that of oxygen
concentration, because the responses are more nonlinear.

So this will be a challenging and interesting area that
can be explored.
4. The constraint controllers developed in this work were

aimed at specifically solving the problems involving the
constraints on manipulated variables. There are various

other constraints present in a real FCC unit. The
control strategy needs to be developed for these
constraints.
192
4• The models developed for the constraint control and
optimization in this work did not involve the pressures

in the two vessels. In a real FCC unit, pressure plays
an important role since it affects the throughput of the
blowers. Therefore, the effect of pressure on the

operation of FCC unit needs to be added in the controller
and optimization models.
6. The optimization was done using Nelder-Mead method.

Since this is an unconstrained optimization method,
constraints were incorporated in the objective function

using penalty functions. This converts the hard
constraints into soft constraints. Therefore, the
optimization should be tried with a better optimization

routine which can handle constraints.
193
BIBLIOGRAPHY
Bromley, J .A. and T.J. Ward. 1981. Fluidized Catalytic Cracker

Control. A Structural Analysis Approach. Ind. Eng. Chem.
Pro. Des. Dev. 20:74.
Choi, J.Y. and R.R. Rhinehart. 1986. Internal Adaptive-Model
Control of Wastewater pH. Paper presented at the 1986
American Control Conference, Minneapolis, MN.
Cutler, C.R. 1983. Dynamic Matrix Control: An Optimal
Multivariable Algorithm with Constraints. Ph.D. Thesis.
Univ. of Houston, U.S.A.
Cutler, C.R., P.H. Gusciora, J.H. McAmis and R.C. Sorensen.
1992. Experiences Applying DMC on a Model IV FCC. Paper
presented at the 1992 AIChE Annual Meeting, Miami, FL.
Economou, C.G., M. Morari and B.O. Palsson. 1986. Internal
Model Control. 5. Extension to Nonlinear Systems.
Ind. Eng. Chem. Proc. Des. Dev. 25:403.
Ford, W.D., I.A. Reineman, I.A. Vaslos and R.J. Fahrig. 1976.
Modeling Catalytic Cracking Regenerators. Paper presented
at the NPRA Annual Meeting, San Antonio, TX.
Garcia, C.E. and M. Morari. 1982. Internal Model Control. 1.
A Unifying Review and Some New Results. Ind. Eng. Chem.
Pro. Des. Dev. 21:308.
Gates, B.C., J.R. Katzer and G.C.A. Schuit. 1979. Chemistry of
Catalytic Processes. McGraw-Hill, New York.
Georgakis, C., R.C. McFarlane, R.C. Reineman and J.F. Bartee.
1990. Dynamic Simulator for a Model IV Fluid Catalytic
Cracking Unit. Paper presented at the 1990 AIChE Annual
Meeting, Chicago, IL.
Gould, L.A., L.B. Evans and B. Kurihara. 1970. Optimal Control
of Fluid Catalytic Cracking Processes. Automatica. 6:695.
Grosdidier, P., A. Mason, A. Aitolahti, P. Heinonen and V.
Vanhamaki. 1993. FCC Unit Reactor-Regenerator Control.
Comp. Chem. Engg. 17:165.
Jacob, S.M., B. Gross, S.E. Voltz and v.w. Weekman, Jr. 1976.
A Lumping and Reaction Scheme for Catalytic Cracking.
AIChE J. 22:701.
Kurihara, H. 1967. Optimal Control of Fluid Catalytic Cracking
Processes. PhD. Thesis. MIT, U.S.A.
194
Lee, W., B.C. Long and v.w. Weekman, Jr. 1973. u.s. Patent
3753893.
Lee, E. and F.R. Groves, Jr. 1985. Mathematical Model of the
Fluidized Bed Catalytic Cracking Plant. Transactions of
the Society for Computer Simulation. Vol. 2. No.3. pp. 219.
Lee, P.L. and G.R. Sullivan. 1988. Generic Model Control.
Comp. Chem. Eng. 12:573.
Pandit, H.G. and R.R. Rhinehart. 1989. Process Model Based
Control of a Fluidized Bed Reactor: A Comparison of Two
Strategies. Instrument Society of America. Advances in
Instrumentation. 44:703.
Paraskos, J.A., Y.T. Shah, J.D. McKinney and N.L. Carr. 1976.
A Kinematic Model for Catalytic Cracking in a Transfer
Line Reactor. Ind. Eng. Chem. Pro. Des. Dev. 15:165.
Parrish, J.R. and C.B. Brosilow. 1988. Nonlinear Inferential
Control. AIChE J. 34:633.
Pohlenz, J.B. 1963. Bow Operational Variables Affect Fluid
Catalytic Cracking. Oil and Gas J. 61(13):124.
Prett, D.M. and R.D. Gillete. 1980. Optimization and
Constrained Multivariable Control of a Catalytic Cracking
Unit. Proceedings of the JACC, San Francisco, CA.
Ramachandran, B., J.B. Riggs and B.B. Beichelheim. 1992.
Nonlinear Plant-Wide Control: Application to a
Supercritical Fluid Extraction Process. Ind. Eng. Chern.
Res. 31:290.
Riggs, J.B. 1988. An Introduction to Numerical Methods for
Chemical Engineers. Texas Tech University Press, Lubbock,
TX.
Riggs, J.B. 1989. Nonlinear Process Model Based Control of a
Column with a Side stream Draw-off. Paper presented at the
1989 AIChE Annual Meeting, San Francisco, CA.
Riggs, J.B. and R.R. Rhinehart. 1991. Two Simple Methods for
On-line Incremental Model Parameterization. Comp. Chem.
Eng. 15:181.
Sculdt, S.B. and F.B. Smith, Jr. 1971. Proceedings of 1971
Joint Automatic Control Conference. pp.270.
voohries, A. 1945. Carbon Formation in Catalytic Cracking.
Ind. Eng. Chem. 37:318.
195
Weekman, v.w., Jr., 1969. Kinetics and Dynamics of Catalytic
Cracking Selectivity in Fixed Bed Reactors. Ind. Eng.
Chem. Pro. Des. Dev. 8:385.
Weekman, v.w., Jr. and D.M. Nace. 1970. Kinetics of Catalytic
Cracking Selectivity in Fixed, Moving, and Fluid Bed
Reactors. AIChE J. 16:397.
~
Weisz, P.W. and R.D. Goodwin. 1963. Combustion of Carbonaceous

Deposits within Porous Catalyst Particles, I. Diffusion-
Controlled Kinetics. Journal of Catalysis. Vol.2. No.5.
pp.397.
Williams, G.L., R.R. Rhinehart and J.B. Riggs. 1990. Inline
Process Model Based Control of Wastewater pH using Dual
Base Injection. Ind. Eng. Chern. Res. 29:1254.
196
APPENDIX A
NOMENCLATURE
= Cross sectional area of lift pipe (8.73 ft 2 )

= Cross sectional area of regenerator (590 ft 2 )
= Cross sectional area of reactor riser ( 9. 6 ft
2
)
= Cross sectional area of standpipe (7 ft 2 )

= Cross sectional area of reactor stripper
(60 ft 2 )
~ndrgc: = Cross sectional area of regenerated catalyst

U-bend ( 3. 7 ft 2 )
~bendec: = Cross sectional area of spent catalyst U-bend

2
(5.2 ft )
= Furnace heat loss parameter (0.15 BTU/s°F)
= Furnace heat loss parameter (200 BTU/s)

= Concentration of carbon monoxide in stack gas
(ppm)
= Weight fraction of hydrogen in coke
(wt H2 /wt coke)
= Concentration of stack gas in regenerator stack

gas (mole %)
= Average heat capacity (BTU/lbmol°F)
cpair = Beat capacity of air (7.08 BTU/lbmol°F)
cpc = Beat capacity of catalyst (0.31 BTU/lb°F)
cpeo = Beat capacity of carbon monoxide

(7.28 BTU/lbmol~)
197
cpco2 Heat capacity of carbon dioxide
(11.0 BTU/lbmol°F)
= Beat capacity of fresh feed liquid
(0.82 BTU/lb-.1°F)
= Beat capacity of fresh feed vapor
( 0. 81 BTU/lb°F)
= Beat capacity of steam (8.62 BTU/lbmol°F)
= Beat capacity of nitrogen (7.22 BTU/lbmol°F)
= Beat capacity of oxygen (7 .62 BTU/~) 1~·--'c. "f"
= Beat capacity of slurry vapor (0.8 BTU/lb°F)

= Weight fraction of coke on regenerated catalyst
(wt coke/wt catalyst)
Wet gas compressor compression ratio
= Weight fraction of coke on spent catalyst
(wt coke/wt catalyst)
= Wet gas production constant
(0.0088438 lbmol/lb feed)
= Wet gas production constant
(0.00004 lbmol/lb feed °F)
= Elevation of regenerator lift air injection
(134 ft)
Eoil inlet = Elevation where oil enters the reactor riser

(124.5 ft)
Eetripper tap = Elevation of reactor stripper tap (130 ft)
Etap = Pressure tap elevation on standpipe (155 ft)
Fair = Air flow rate into regenerator (lbmol/s)
198
= Effect of feed type on coke production
= Air lift compressor inlet suction flow at base
conditions (ICFM)
F coke = Production of coke in reactor riser (lb/s)
FH = Burning rate of hydrogen (lb/s)
fof = Overflow factor
fpp (X) = Nonlinear valve flow rate function
frgc = Force exerted by regenerated catalyst (lb£)
Frgc = Flow rate of regenerated catalyst (lb/s)
fiSC = Force exerted by spent catalyst (lb£)
FISC = Flow rate of spent catalyst (lb/s)
Frsg = Stack gas flow (lbmol/s)
Frsp = Flow into standpipe (lb/s)
Frsucn 1 comb = Combustion air blower inlet suction flow ( ICFM)
F eucn 1 lift = Lift air blower inlet suction flow (ICFM)
F sucn 1 wg = Wet gas compressor suction flow (ICFM)
F surge 1 comb = Combustion air blower surge flow (ICFM)
F eurge 1 lift = Lift air blower surge flow (ICFM)
Frsurge 1 wg = Wet gas compressor surge flow (ICFM)
FT = Air flow rate into regenerator (lb/s)
fubendrgc = Regenerated catalyst friction factor

2
( 23 lb s/ft )
2
fubendrsc = Spent catalyst friction factor (47 lb s/ft )
= Flow through combustion air blower suction
valve (V6 ) (lb/s)

= Flow through combustion air blower vent valve
199
(V7 ) ( lb/s)
= Flow through lift air blower vent valve (V8 )
= Flow through wet gas compressor suction valve
(V11 ) (lbmol/s)
= Flow through wet gas flare valve (V 12 ) (lbmol/s)
= Flow through wet gas anti-surge valve (V13 )
(lbmol/s)
= Wet gas production in reactor (lbmol/s)
= Flow of fresh feed through reactor riser ( lb/ s)
= Flow of slurry to reactor riser (lb/s)
= Flow of fuel to furnace (SCF/s)
= Setpoint to furnace fuel flow controller

(SCF/s)
= Combustion air blower throughput (lb/s)
= Combustion air flow to the regenerator (lb/s)
= Lift air blower throughput (lb/s)
= Lift air flow to the regenerator (lb/s)
= Spill air flow to the regenerator (lb/s)
= Wet gas flow to the vapor recovery unit
(lbmol/s)
2
g = Acceleration due to gravity (32.2 ft lb./s lbf)
= Height of lift pipe (34 ft)
= Height of reactor riser (60 ft)
= Height of regenerator standpipe (20 ft)
= Wet gas compressor head
= Average ratio of specific heats (1.39)
200
= Combustion air blower discharge pipe flow
resistance factor (62.94 lb/s psia 0 • 5 )
= Lift air blower discharge pipe flow resistance
factor (5 lb/s psia0 • 5 )
= Reaction rate constant for reaction I
= Reaction rate constant for reaction II
= Reaction rate constant for reaction III
Combustion air blower suction valve flow rating
( 250 lb/ s psiao.s)
= Combustion air blower vent valve flow rating
(25 lb/s psia 0 • 5 )
= Lift air blower vent valve flow rating
( 10 lb/s psia 0 • 5 )
= Lift air blower spill valve flow rating

( 10 lb/s psia 0 • 5 )
= Wet gas compressor suction valve flow rating

( 1. 5 lbmol/ s psia0 • 5 )
= Wet gas flare valve flow rating
(0.5 lbmol/s psia 0 • 5 )
= Wet gas compressor anti-surge valve flow rating

0 5
(0.01 lbmol/s psia • )
= Regenerator stack gas valve flow rating

(0.01 lbmol/s psia 0 • 5 )
Level of catalyst in standpipe (ft)
Lubendrqc
= Length of regenerated catalyst U-bend (56 ft)
Lubendec
= Length of spent catalyst U-bend (56 ft)
201
M = Polytropic exponent
= Effective heat capacity of riser vessel and
catalyst (10,000 BTU/°F)
Flow rate of entrained catalyst from dense bed
into dilute phase (lb/s)
= Effective heat capacity of regenerator mass
( 200,000 BTU/°F)
= Inertial mass of regenerated catalyst ( lbts 2 I ft)
= Inertial mass of spent catalyst (lbfs 2 /ft)
n = Amount of gas (lbmol)
patm Atmospheric pressure (14.7 psia)

= Combustion air blower base discharge pressure
(psia)
pbaae,d = Lift air blower base discharge pressure (psia)
pblp = Pressure at bottom of lift pipe (psi)
prb Pressure at bottom of reactor riser (psi)
prgb = Pressure at bottom of regenerator (psi)
pvru = Discharge pressure of wet gas compressor to

vapor recovery unit (101 psia)
= Combustion air blower suction pressure (psia)
= Combustion air blower discharge pressure ( psia)

Lift air blower discharge pressure (psia)
= Reactor pressure (psia)
= Reactor fractionator pressure (psia)

= Regenerator pressure (psia)
= Wet gas compressor suction pressure (psia)
202
= Enthalpy of incoming air (BTU/s)
Total heat of burning carbon (BTU/s)
= Enthalpy of catalyst out of reactor riser
(BTU/s)
= Heat generated from cracking (BTU/s)
= Total heat lost from regenerator to environment
(556 BTU/s)
= Beat required to bring fresh feed to reactor
riser temperature (BTU/s)
= Enthalpy of outgoing stack gas (BTU/s)
= Beat required to raise temperature of slurry
from 500°F(liquid) to 1000°F(vapor) (309 BTU/s)
= Enthalpy of incoming hydrogen (BTU/s)
Qin = Enthalpy into regenerator (BTU/s)
Qloee = Beat loss from furnace (BTU/s)
Qout = Enthalpy out of regenerator (BTU/s)
Qrgc = Enthalpy of regenerated catalyst (BTU/s)
Qrin = Enthalpy into reactor (BTU/s)
= Enthalpy out of reactor (BTU/s)

= Enthalpy of spent catalyst (BTU/s)
= Beat required to bring slurry to reactor riser

temperature (BTU/s)
Heat required to raise temperature of fresh
feed from 700°F(liquid) to 1000°F(vapor)
(412 BTU/s)
3 0
R = Universal gas constant ( 10.73 ft psi/lbmol R)
203
= Actual speed of lift air blower (RPM)
smax = Maximum speed of the lift air blower ( 6100 RPM)
a
S ain
a = Minimum speed of the lift air blower ( 5000 RPM)
= Base speed of the lift air blower (5950 RPM)

t = Time (s)
Tal.r = Temperature of air entering regenerator
(270 °F)
Tatm = Atmospheric temperature (75 °F)
Tbaae = Base temperature (1100 °F)

= Base temperature of reactor fresh feed ( 700 °F)
T ccmb,d = Combustion air blower discharge temperature
( 190 °F)
Tcyc = Regenerator stack gas temperature at cyclone
= Regenerator bed and cyclone temperature
difference (°F)
Tlift,d = Lift air blower discharge temperature (225 °F)
Tlm = Furnace log mean temperature difference (°F)
= Temperature of reactor riser (°F)
Tref = Base temperature of reactor riser energy
balance ( 999 °F)
Treg = Temperature of regenerator bed (°F)
= Temperature of spent catalyst entering
regenera t or ( OF)
= Temperature of fresh feed entering regenerator
204
= Temperature of fresh feed entering reactor
riser (°F)
T 2, •• = Steady state furnace outlet temperature (°F)
= Furnace firebox temperature (°F)
= Furnace overall heat transfer coefficient
(25 BTU/s)
vair,lift = Velocity of air in lift pipe (ft/s)
V cat,lift = Velocity of catalyst in lift pipe (ft/s)
vcomb,d = Combustion air blower discharge system volume

( 1000 ft 3 )
vlift = Manipulated variable for lift air blower steam

valve ( 0-1)
Lift air blower discharge system volume
(200 ft 3 )
= Regenerator volume occupied by gas (ft 3 )
vrgc = Velocity of regenerated catalyst (ft/s)
vri• = Volumetric flow rate in reactor riser ( ft 3 /s)

= Superficial velocity in regenerator (ft/s)
= Velocity of spent catalyst (ft/s)
v.up = Slip velocity (2.2 ft/s)

= Combustion air blower suction valve position
(0-1)
= Combustion air blower vent valve position ( 0-1)
= Lift air blower vent valve position (0-1)
= Spill air valve position (0-1)

= Wet gas compressor suction valve position ( 0-1)
205
= Wet gas flare valve position (0-1)
= Wet gas compressor vent valve position (0-1)
= Stack gas valve position (0-1)
= Inventory of carbon in regenerator (lb)
= Inventory of carbon in regenerator (lb)
wreg = Inventory of catalyst in regenerator (lb)
wrie = Inventory of catalyst in reactor riser (lb)
= Inventory of catalyst in regenerator standpipe
(lb)
WHSV = Weight hourly space velocity
(lb oil/(lb cat/hr))
= Molar ratio of CO to air (moles CO/moles air)
= Molar ratio of CO to air in stack gas

(moles CO/moles air)
= Molar ratio of C02 to air (moles C02 /moles air)
= Molar ratio of C02 to air in stack gas
(moles C0 2 /moles air)
= Molar ratio of N2 to air (moles N2 /moles air)
= Molar ratio of 0 2 to air (moles 0 2 /moles air)
= Molar ratio of 0 2 to air in stack gas
(moles 0 2 /moles air)

z Vertical position in regenerator (ft)
zbed = Regenerator dense bed height (ft)
= Height of cyclone inlet (45 ft)

= Height of lift pipe discharge (11 ft)
= Standpipe exit height from bottom of
206
regenerator (13 ft)
z top = 0 2 , CO measurement point (47 ft)
GREEK SYMBOLS
6z = Dirac delta function
AB£ = Heat of cracking (BTU/lb)
ABtu = Beat of combustion of furnace fuel
(1000 BTU/SCF)
= Beat of combustion of hydrogen (60,960 BTU/lb)
Heat of formation of CO (BTU/lbmol CO produced)
= Beat of formation of C02 (BTU/lbmol C02

produced)
~p frac = Pressure drop across reactor main fractionator
(9.5 psia)
~prgc = Pressure drop from bottom of standpipe to oil
inlet elevation of reactor riser (psi)
= Differential pressure between reactor .and
regenerator (psi)
l\Psc = Pressure drop from bottom of reactor to lift
air injection elevation (psi)
~Tstripper = Temperature drop across reactor stripper ( 35°F)
ee = Effective void fraction in regenerator dense
phase bed
= Apparent void fraction in regenerator dense
phase bed
n
p
= Polytropic efficiency (1)
207
Pairg = Density of air at regenerator conditions
3
( lb/ft )
= Volume fraction of catalyst
= Density of catalyst in U-bend and regenerator

3
standpipe (45 lb/ft )
Pc,dilute = Density of catalyst in the dilute phase

3
(lb/ft )
3
Pc,denae = Density of catalyst in the dense bed (lb/ft )
= Density of exit gas (lbmol/ft 3 )

3
plift = Density of catalyst in lift pipe (lb/ft )
3
Ppart = Settled density of catalyst (68 lb/ft )
Pria = Average density of material in reactor riser

3
(lb/ft )
= Vapor density at reactor riser conditions

3
(0.57 lb/ft )
t:fb
Furnace firebox time constant (200 s)
= Riser fill time (40 s)
= Furnace time constant (60 s)

Catalyst residence time in riser (min)
Effective coke factor for gas oil feed
208
APPENDIX B
INITIAL CONDITIONS FOR DYNAMIC SIMULATION
Table B.1 : Initial Conditions for

State Variables
State Variable Initial Value
Treg 1271.168 OF
crgc 8. 537 X 10-4 lb/lb
n 246.027 lbmol
we 1292.407 lb
w.p 3555.173 lb
wreg 273815.2 lb
Plitt 3.27971 lb/ft 3

wr 101298.377 lb
Tr 994.067 °F
c.c 7 • 829 X 10-3 lb/lb

pl 14.63286 psia
p2 35.1974 ps~a
.
p3 40.3669 ps~a
.
p5 23.51966 psia
p6 29.64251
.
ps~a
p7 22.60185 .
ps.1.a
T2 667.261 OF
T] 1607.55 OF
209
Table B.2 : Initial Setpoints for
Regulatory Controllers
Controller Setpoints Initial Value
••t 131.25 lb/s

Foil
F eet 34.00 SCF/s
•
F ..t 75.46545 lb/s
T
p ••t 33.01966 psl.a

.
•
F ..t 760.0 lb/s
•
210
Table B.3 : Initial Valve Positions
Valve Initial Value
vs 1.0000
v, 0.0000
ve 0.0000
v, 0.0000
vu 0.9500
v12 0.0000
v13 0.0000
vu 0.6120
Van 0.4238
211
APPENDIX C
STEADY-STATE DATA FOR PARAMETERIZATION
Table C.1 : Steady-State Feed Pre-Beater Data
FOYi Fs T T T
(lb s) (SCF/s) (oF) (oF) (oF)
127.0 34.0 460.9 665.67 1606.84

128.0 34.0 460.9 664.09 1606.15
129.0 34.0 460.9 662.55 1605.46
126.0 34.0 460.9 667.26 1607.55
125.0 34.0 460.9 668.883 1608.27
124.0 33.0 460.9 665.51 1581.50
123.0 36.0 460.9 682.16 1663.88
122.0 37.0 460.9 688.88 1691.31
128.0 36.0 460.9 673.68 1660.15
129.0 32.0 460.9 652.83 1550.30
122.0 33.0 470.9 678.47 1591.18
124.0 36.0 450.9 . 670.76 1655.04
126.0 32.0 440.9 639.95 1535.81
121.0 36.0 480.9 705.03 1681.60
128.0 31.0 465.9 654.18 1527.07
127.0 36.0 455.9 670.49 1656.84
126.0 35.0 460.9 672.15 1634.72
126.0 38.0 460.9 686.51 1714.52
126.0 33.0 460.9 662.32 1580.08

126.0 37.0 460.9 681.78 1688.21
212
Table C.2 : Reactor-Regenerator Steady-State Data
Treg Tr c.c crgc 02 Conv

(Mo i. 9%) (Wt %)
1246.4 986.7 0.7547 0.0923 2.49 66.92
1255.0 989.3 0.7646 0.0900 2.15 67.69
1257.7 990.1 0.7678 0.0894 2.04 67.92
1260.3 990.9 0.7709 0.0888 1.94 68.15
1262.8 991.6 0.7740 0.0883 1.84 68.37
1265.2 992.3 0.7770 0.0879 1.74 68.57
1267.6 993.0 0.7800 0.0876 1.65 68.77
1267.7 993.1 0.7800 0.0875 1.65 68.78
1268.5 993.3 0.7811 0.0874 1.61 68.84
1269.8 993.7 0.7829 0.0873 1.56 68.95
1271.9 994.3 0.7857 0.0870 1.48 69.12
1272.82 994.6 0.7868 0.0869 1.44 69.20
1274.06 994.9 0.7885 0.0868 1.39 69.30
1276.05 995.5 0.7912 0.0867 1.31 69.55
1279.80 996.6 0.7964 0.0866 1.17 69.75
1283.22 997.6 0.8014 0.0866 1.03 70.01
1281.55 997.1 0.7990 0.0866 1.10 69.88
1264.26 992.0 0.7758 0.0881 1.78 68.49
213
APPENDIX D
DYNAMIC RESPONSES FROM ORIGINAL SIMULATOR
214
~ 1300.-------~------~------~------~------------~
Ol
Q)
-
"0
Q)
J.-1
....,::J
cU
J..l
Q)
~ 1280
Q)
E-t
J..l
....,0
aj
J.-1
Q) i i i i
;Ol !: I
:
l
:
I
:
~ ~ i ;
Q) 1260+--------+--------~------+-------~------~------~
~ 0 25 50 75 100 125 150
Time (Minutes) ·
Figure D.l Dynamic Response of Regenerator Temperature

Formation Factor
215
1004
l'z.c
0'
(1)
-cv
"0
~
::s
~
cu
~ 999
i
(1)
0..
~
cv
E-t
~
0
~
u
cu
cv
c:t::
l j
994+--------+--------~------+-------~-------4------~
0 25 50 75 100 125 150
Time (Minutes)
Figure D.2 Dynamic Response of Reactor Temperature

Formation Factor
216
..Q 0 .0084
...-t
.........
..Q
-
r-i
.....,
Ul
:>.,
...-t
CIS
.....,
--r------ :---·--·---
CIS i
u
---~-------~--------r-
0 .008 -
.....,
~
Q)
p.
{/) ! : . : z
c::
0 1.• I
'
!
:
I
t
I
I
~
0
..Q
H
u
CIS
0.0076
0
I
25
i
50
!
75
i
100
I
125 150
Time (Minutes)
Figure D.3 Dynamic Response of Carbon Concentration on

Spent Catalyst for 0.05 Step Change in Fresh
Feed Coke Formation Factor
217
.Q
r-4
:!i 0 .0011....-----.------,-------,-------,-------,-------,
-
r-4
~
>,
i
i
r-4 i
«J i
~ 0.001 : : : ' I
u
"0
Q)
+l
«J
~~ 0 .0009 -
I - - ---·r.
-----·r·------- ---------------r-----------·--r----·------------l_.-----------------
~ I! I!
r:: !
j
j
~
0 : :
~
0
l
0 . 0008+-------+=-------+
l
= -------r------~------+-----~
-e 0 25 50 75 100 125 150
co Time (Minutes)
u
Figure D.4 Dynamic Response of Carbon Concentration on
.
Regenerated Catalyst for 0.05 Step Change l.n
Fresh Feed Coke Formation Factor
218
-
~
r:::
2~-------,--------~------~~------~--------~------~
0 l
-~ I
~u Ii
a
~ 1
---- --r -------r - - T -- -r-------r--------
1
0'\
:>.
~
0
C1l
ell
(!)
Q)
~ O+---------+i--------~--------4---------+---------~------~
~ 0 25 50 75 100 125 150
Time (Minutes)
Figure D.S Dynamic Response of Flue Gas Oxygen

Concentration for 0.05 Step Change
in Fresh Feed Coke Formation Factor
219
30
......
m
- p.
<J)
~
:::1
m
m
Q,)
~
P-1 29.75
~
0
-4-J
cu
~
Q,)
~
(])
Ol
Q,)
~
i i ~ i
29.5+--------r-------+------~------~--------r-----~
0 25 50 75 100 125 150
Time (Minutes)
Figure D.6 Dynamic Response of Regenerator Pressure for

0.05 Step Change in Fresh Feed Coke Formation
Factor
220

Control and Optimization of Fluidized Catalytic Cracking Process

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Control and Optimization of Fluidized Catalytic Cracking Process

Caricato da

Copyright:

Formati disponibili

CONTROL AND OPTIMIZATION OF FLUIDIZED CATALYTIC

Submitted to the Graduate Faculty

I wish to express my sincere appreciation to Prof. James

extend my sincere thanks to my other committee members, Dr.

1.3. Scope of Work 4

2.1. Fluidized Catalytic Cracking Process

2.1.2. Regeneration of Catalyst 9

2.1.3.1. Dynamic Simulation of FCC Unit 11

2.2.1. Linear Model Based Control 18

3. PROCESS DESCRIPTION AND DYNAMIC SIMULATION 24

3.3.1. Amoco Model 27

4.3.4. Reactor Temperature Control 118

5. CONSTRAINT CONTROL 121

6.2.1. Gasoline Yield Model 139

6.2.3. Regenerator Oxygen Balance 143

6•. 3. Optimization Procedure 144

7. RESULTS AND DISCUSSION 146

a. CONCLUSIONS AND RECOMMENDATIONS 190

B. INITIAL CONDITIONS FOR DYNAMIC SIMULATION 209

c. STEADY-STATE DATA FOR PARAMETERIZATION 212

D. DYNAMIC RESPONSES FROM ORIGINAL SIMULATOR 214

7.1 Tuning Constants for Basic Regulatory

7.2 Tuning Constants for Upper Level

2.1 Schematic of Conventional Control Scheme 22

3.12 Open-Loop Response of Reactor Temperature

3.16 Open-Loop Response of Regenerator Temperature

3.20 Open-Loop Response of Flue Gas Oxygen

3.21 Open-Loop Response of Reactor Temperature

3.25 Open-Loop Response of Flue Gas Oxygen

3.29 Open-Loop Response of Carbon on Regenerated

3.31 Open-Loop Response of Reactor Temperature for

3.34 Open-Loop Response of Reactor Temperature for

3.37 Open-Loop Response of Flue Gas Oxygen

3.38 Open-Loop Response of Regenerator Temperature

7.1 Unconstrained Control Response of Reactor

7.2 Unconstrained Control Response of Regenerator

7.3 Unconstrained Control Response of Flue Gas

7.4 Unconstrained Control Response of Reactor

7.5 Unconstrained Control Response of Regenerator

7.6 Unconstrained Control Response of Flue Gas

1.1 Unconstrained Control Response of Reactor

The Fluid Catalytic Cracking Unit (FCCU) receives

lighter and more valuable components. After further

1.1. Nature of the FCCU Process

Generally the FCC process can be divided into three major

stream of regenerated catalyst. The reaction occurs over a

short contact time in a riser before the catalyst and reacted

1 ike gasoline. The degree of gas oil conversion and coke

yield in the riser are a complex function of the reactor

regenerated by burning off the carbon before circulating back

to the feed injector. The temperature in the reactor and the

regenerator will remain stable when the heat balance between

regenerator. Similarly, the carbon balance between the two

vessels should be satisfied, i.e., the carbon generated in the

reactor riser must be burned in the regenerator to keep the

carbon concentration on the catalyst as low as possible.

Thus, the effect of operating conditions on conversion and

yields in the reactor can be grouped into two simple terms.

"catalyst severity," which depends on the catalyst-to-oil

ratio, catalyst contact time an& catalyst activity. The unit

operator can measure the reactor temperature, but he has no

direct measure of the vital catalyst severity.

Fluidized catalytic Cracker is one of the most difficult